[{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"C is the union of the sets of constituents suggested by the parsers. r(c) is a binary function returning t (for true) precisely when the constituent c E C should be included in the hypothesis.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"We compare three hash tables: our probing implementation, GCC\u2019s hash set, and Boost\u2019s8 unordered.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"The last affix in the list is the nominal plural f, men0.20 In the table are the (typical) classes of words to which the affix attaches, the number found in the test corpus by the method, the number correct (with a precision measure), and the number missed (with a recall measure).","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Nonetheless there is no alternative to Reiche\u00e2\u0080\u0099s plan.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Their default variant implements a forward trie, in which words are looked up in their natural left-to-right order.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"4.2 Global Features.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"The algorithm works due to the fact that not all permutations of cities have to be considered explicitly.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"And if one is interested in TIS, one would probably consider the single orthographic word ACL to consist of three phonological words-lei s'i d\/-corresponding to the pronunciation of each of the letters in the acronym.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"As a first step towards modeling transliterated names, we have collected all hanzi occurring more than once in the roughly 750 foreign names in our dictionary, and we estimate the probabil\u00c2\u00ad ity of occurrence of each hanzi in a transliteration (pTN(hanzi;)) using the maximum likelihood estimate.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"This revealed interesting clues about the properties of automatic and manual scoring.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Since the segmentation corresponds to the sequence of words that has the lowest summed unigram cost, the segmenter under discussion here is a zeroth-order model.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"Our first model is GTplain, a PCFG learned from the treebank after removing all functional features from the syntactic categories.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"The similar explanation applies to the link to the \u00e2\u0080\u009cstake\u00e2\u0080\u009d set.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Ex: Mr. Cristiani, president of the country ...","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"In our work, we demonstrate that using a simple na\u00c2\u00a8\u00c4\u00b1veBayes approach also yields substantial performance gains, without the associated training complexity.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"The best answer to this is: many research labs have very competitive systems whose performance is hard to tell apart.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"The first issue relates to the completeness of the base lexicon.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"Finally, we show that in application settings, the absence of gold segmentation lowers parsing performance by 2\u00e2\u0080\u00935% F1.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Using RandLM and the documented settings (8-bit values and 1 256 false-positive probability), we built a stupid backoff model on the same data as in Section 5.2.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"The second row is the accuracy of the best of the three parsers.'","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"He used an additional maximum entropy classifier that tries to correct mistakes by using reference resolution.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Surprisingly, this effect is much less obvious for out-of-domain test data.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"On the MUC6 data, Bikel et al.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Annotation consistency is important in any supervised learning task.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"31 75.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"We use double-circles to indicate the space-delimited token boundaries.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"We refer to (T , W ) as the lexicon of a language and \u00cf\u0088 for the parameters for their generation; \u00cf\u0088 depends on a single hyperparameter \u00ce\u00b2.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"This essentially means that their model does not result in a true joint inference, as executions for different tasks involve tuning a parameter separately.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"For example, in Information Retrieval (IR), we have to match a user\u00e2\u0080\u0099s query to the expressions in the desired documents, while in Question Answering (QA), we have to find the answer to the user\u00e2\u0080\u0099s question even if the formulation of the answer in the document is different from the question.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"We thank members of the MIT NLP group for their suggestions and comments.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"As noted, this sentence consists of four words, namely B X ri4wen2 'Japanese,' :\u00c2\u00a5, zhanglyu2 'octopus\/ :&P:l zen3me0 'how,' and IDt shuol 'say.'","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"For one thing, it is not clear who is to receive settlements or what should happen in case not enough teachers accept the offer of early retirement.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"For the seen word ir, 'gen\u00c2\u00ad erals,' there is an c:NC transduction from to the node preceding ir,; this arc has cost cost( f,) - cost(unseen(f,)), so that the cost of the whole path is the desired cost( f,).","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"This technique was introduced by Clarkson and Rosenfeld (1997) and is also implemented by IRSTLM and BerkeleyLM\u2019s compressed option.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"Table 2 shows BABAR\u00e2\u0080\u0099s performance.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"We extend the Matsoukas et al approach in several ways.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"Across all languages, high performance can be attained by selecting a single tag per word type.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"As expected, the vanilla HMM trained with EM performs the worst.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Table 2 shows single-threaded results, mostly for comparison to IRSTLM, and Table 3 shows multi-threaded results.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"So we decided to use semantic class information only to rule out candidates.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"For Arabic we M o d e l S y s t e m L e n g t h L e a f A n c e s t o r Co rpu s Sent Exact E v a l b L P LR F1 T a g % B a s e l i n e 7 0 St an for d (v 1.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"the number of permutations carried out for the word reordering is given.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"By aggregating the POS labels of the English tokens to types, we can generate label distributions for the English vertices.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Using this heuristic, BABAR identifies existential definite NPs in the training corpus using our previous learning algorithm (Bean and Riloff, 1999) and resolves all occurrences of the same existential NP with each another.1 2.1.2 Syntactic Seeding BABAR also uses syntactic heuristics to identify anaphors and antecedents that can be easily resolved.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Annotators argued for the importance of having correct and even multiple references.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"In the next section we present an alternative approach that builds two classifiers while attempting to satisfy the above constraints as much as possible.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"We also report word type level accuracy, the fraction of word types assigned their majority tag (where the mapping between model state and tag is determined by greedy one-to-one mapping discussed above).5 For each language, we aggregate results in the following way: First, for each hyperparameter setting, evaluate three variants: The first model (1TW) only 4 Typically, the performance stabilizes after only 10 itera-.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"The table makes clear that enriching our grammar improves the syntactic performance as well as morphological disambiguation (segmentation and POS tagging) accuracy.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"In the maximum entropy framework, there is no such constraint.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"The 2nd block contains the IR system, which was tuned by selecting text in multiples of the size of the EMEA training corpus, according to dev set performance.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"The PROBING model can perform optimistic searches by jumping to any n-gram without needing state and without any additional memory.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"(student+plural) 'students,' which is derived by the affixation of the plural affix f, menD to the nounxue2shengl.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"We define the path set of a tree 1 as the set of strings that label a path from the root to frontier of 7.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Topicalization of NP subjects in SVO configurations causes confusion with VO (pro-drop).","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Consequently, all three parsers prefer the nominal reading.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Mohri [1995]) shows promise for improving this situation.","label":0},{"source_text":"There is no global pruning.","target_text":"The baseline alignment model does not permit that a source word is aligned to two or more target words, e.g. for the translation direction from German toEnglish, the German compound noun 'Zahnarztter min' causes problems, because it must be translated by the two target words dentist's appointment.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Among these are words derived by various productive processes, including: 1.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"Comparison with state-of-the-art taggers For comparison we consider two unsupervised tag- gers: the HMM with log-linear features of Berg- Kirkpatrick et al.","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"To explore this tradeoff, we have performed experiments with three different encoding schemes (plus a baseline), which are described schematically in Table 1.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"This is less than the 694 judgements 2004 DARPA\/NIST evaluation, or the 532 judgements in the 2005 DARPA\/NIST evaluation.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"This Good\u00c2\u00ad Turing estimate of p(unseen(f,) If,) can then be used in the normal way to define the probability of finding a novel instance of a construction in ir, in a text: p(unseen(f,)) = p(unseen(f,) I f,) p(fn Here p(ir,) is just the probability of any construction in ft as estimated from the frequency of such constructions in the corpus.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"JA DE G O L D G R AS S SI C K NE SS DE AT H R A T 14.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"The price shocked her.), or same-kind (e.g., Her health insurance paid for the hospital fees, but the automobile insurance did not cover the repair.).","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"J:j:l :zhongl :0.0 ;m,Jlong4 :0.0 (mHHaryg9tltHBI) \u00c2\u00a3: _ADV: 5.98 \u00c2\u00a5 :hua2:o.o E :_NC: 4.41 :mln2:o.o mm : guo2 : 0.0 (RopubllcofChlna) .....,.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"If two nouns have mutually exclusive semantic classes, then they cannot be coreferent.","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"This may seem surprising, given the experiments reported in section 4, but the explanation is probably that the non-projective dependencies that can be recovered at all are of the simple kind that only requires a single lift, where the encoding of path information is often redundant.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"Like the string languages of MCTAG's, the complexity of the path set increases as the cardinality of the elementary tee sets increases, though both the string languages and path sets will always be semilinear.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"We respond to this on the one hand with a format for its underspecification (see 2.4) and on the other hand with an additional level of annotation that attends only to connectives and their scopes (see 2.5), which is intended as an intermediate step on the long road towards a systematic and objective treatment of rhetorical structure.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"We can 5 Recall that precision is defined to be the number of correct hits divided by the total number of items.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"att.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"Since our goal is to perform well under these measures we will similarly treat constituents as the minimal substructures for combination.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Fortunately, we were able to obtain a copy of the full set of sentences from Chang et al. on which Wang, Li, and Chang tested their system, along with the output of their system.18 In what follows we will discuss all cases from this set where our performance on names differs from that of Wang, Li, and Chang.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"99 94.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Instead, we extend the variation n-gram method of Dickinson (2005) to compare annotation error rates in the WSJ and ATB.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"Parameter Component As in the standard Bayesian HMM (Goldwater and Griffiths, 2007), all distributions are independently drawn from symmetric Dirichlet distributions: 2 Note that t and w denote tag and word sequences respectively, rather than individual tokens or tags.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"It is not immediately obvious how to formulate an equivalent to equation (1) for an adapted TM, because there is no well-defined objective for learning TMs from parallel corpora.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"For each case- frame, BABAR collects the head nouns of noun phrases that were extracted by the caseframe in the training corpus.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"19 We note that it is not always clear in Wang, Li, and Chang's examples which segmented words.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"From the definition of TAG's, it follows that the choice of adjunction is not dependent on the history of the derivation.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"There are simply not enough votes remaining to allow any of the crossing structures to enter the hypothesized constituent set.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"If one system is perfect, another has slight flaws and the third more flaws, a judge is inclined to hand out judgements of 5, 4, and 3.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"A search restriction especially useful for the translation direction from German to English is presented.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"Again, this deserves further investigation.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Table 3: Dev set frequencies for the two most significant discourse markers in Arabic are skewed toward analysis as a conjunction.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"For example, kidnapping victims should be extracted from the subject of the verb \u00e2\u0080\u009ckidnapped\u00e2\u0080\u009d when it occurs in the passive voice (the shorthand representation of this pattern would be \u00e2\u0080\u009c<subject> were kidnapped\u00e2\u0080\u009d).","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"\u00e2\u0080\u009cThe gun\u00e2\u0080\u009d will be extracted by the caseframe \u00e2\u0080\u009cfired <patient>\u00e2\u0080\u009d.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"(Riloff and Shepherd 97) describe a bootstrapping approach for acquiring nouns in particular categories (such as &quot;vehicle&quot; or &quot;weapon&quot; categories).","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"At present, the \u00e2\u0080\u0098Potsdam Commentary Corpus\u00e2\u0080\u0099 (henceforth \u00e2\u0080\u0098PCC\u00e2\u0080\u0099 for short) consists of 170 commentaries from Ma\u00c2\u00a8rkische Allgemeine Zeitung, a German regional daily.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"The ATB annotation guidelines specify that proper nouns should be specified with a flat NP (a).","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"Using a wide-coverage morphological analyzer based on (Itai et al., 2006) should cater for a better coverage, and incorporating lexical probabilities learned from a big (unannotated) corpus (cf.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"Their theoretical finding is simply stated: classification error rate decreases toward the noise rate exponentially in the number of independent, accurate classifiers.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective treebanks is often neglected because of the relative scarcity of problematic constructions.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"(2009b) evaluated the Bikel parser using the same ATB split, but only reported dev set results with gold POS tags for sentences of length \u00e2\u0089\u00a4 40.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"For each pair of judges consider one judge as the standard,.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Like in the co-reference annotation, G\u00c2\u00a8otze\u00e2\u0080\u0099s proposal has been applied by two annotators to the core corpus but it has not been systematically evaluated yet.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"In Section 2, we brie y review our approach to statistical machine translation.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"Our \u201cProjection\u201d baseline is able to benefit from the bilingual information and greatly improves upon the monolingual baselines, but falls short of the \u201cNo LP\u201d model by 2.5% on an average.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"Most of these groups follow a phrase-based statistical approach to machine translation.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"The second model (+PRIOR) utilizes the independent prior over type-level tag assignments P (T |\u00cf\u0088).","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"With each iteration more examples are assigned labels by both classifiers, while a high level of agreement (> 94%) is maintained between them.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"For each feature type f and tag t, a multinomial \u00cf\u0088tf is drawn from a symmetric Dirichlet distribution with concentration parameter \u00ce\u00b2.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Diacritics can also be used to specify grammatical relations such as case and gender.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"Unsupervised Models for Named Entity Classification Collins","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"4.5 Transliterations of Foreign Words.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"The developers aimed to reduce memory consumption at the expense of time.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"For previously unseen hanzi in given names, Chang et al. assign a uniform small cost; but we know that some unseen hanzi are merely acci\u00c2\u00ad dentally missing, whereas others are missing for a reason-for example, because they have a bad connotation.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Another option is the closedsource data structures from Sheffield (Guthrie and Hepple, 2010).","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"One is the accuracy within a set of phrases which share the same keyword; the other is the accuracy of links.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"1 The apparent difficulty of adapting constituency models to non-configurational languages has been one motivation for dependency representations (Hajic\u00cb\u0087 and Zema\u00c2\u00b4nek, 2004; Habash and Roth, 2009).","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"Taking \/V, to be the number of examples an algorithm classified correctly (where all gold standard items labeled noise were counted as being incorrect), we calculated two measures of accuracy: See Tab.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Fortunately, we were able to obtain a copy of the full set of sentences from Chang et al. on which Wang, Li, and Chang tested their system, along with the output of their system.18 In what follows we will discuss all cases from this set where our performance on names differs from that of Wang, Li, and Chang.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"We adopted this state-of-the-art model because it makes it easy to experiment with various ways of incorporating our novel constraint feature into the log-linear emission model.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Following the system devised under the Qing emperor Kang Xi, hanzi have traditionally been classified according to a set of approximately 200 semantic radicals; members of a radical class share a particular structural component, and often also share a common meaning (hence the term 'semantic').","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"am 11.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"There are thus some very good reasons why segmentation into words is an important task.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"Our derivation is slightly different from the one presented in (Schapire and Singer 98) as we restrict at to be positive.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Participants and other volunteers contributed about 180 hours of labor in the manual evaluation.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"Terrorism systems must distinguish between people who perpetrate a crime and people who are victims of a crime.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"Moving beyond directly related work, major themes in SMT adaptation include the IR (Hildebrand et al., 2005; L\u00a8u et al., 2007; Zhao et al., 2004) and mixture (Finch and Sumita, 2008; Foster and Kuhn, 2007; Koehn and Schroeder, 2007; L\u00a8u et al., 2007) approaches for LMs and TMs described above, as well as methods for exploiting monolingual in-domain text, typically by translating it automatically and then performing self training (Bertoldi and Federico, 2009; Ueffing et al., 2007; Schwenk and Senellart, 2009).","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Finally, quite a few hanzi are homographs, meaning that they may be pronounced in several different ways, and in extreme cases apparently represent different morphemes: The prenominal modifi\u00c2\u00ad cation marker eg deO is presumably a different morpheme from the second morpheme of \u00c2\u00a7eg mu4di4, even though they are written the same way.4 The second point, which will be relevant in the discussion of personal names in Section 4.4, relates to the internal structure of hanzi.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"2 56.2 32.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"This makes memory usage comparable to our PROBING model.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"First, we evaluated BABAR using only the seven general knowledge sources.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"See Table 2 for the tag set size of other languages.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"Schapire and Singer show that the training error is bounded above by Thus, in order to greedily minimize an upper bound on training error, on each iteration we should search for the weak hypothesis ht and the weight at that minimize Z.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Tables 4 and 5 show BABAR\u00e2\u0080\u0099s performance when just one contextual role knowledge source is used at a time.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"The points enumerated above are particularly related to ITS, but analogous arguments can easily be given for other applications; see for example Wu and Tseng's (1993) discussion of the role of segmentation in information retrieval.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"The NIST06 and NIST08 evaluation sets were used for testing.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"Similar behavior is observed when adding features.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Global context from the whole document is available and can be exploited in a natural manner with a maximum entropy classifier.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"The choice between different actions is in general nondeterministic, and the parser relies on a memorybased classifier, trained on treebank data, to predict the next action based on features of the current parser configuration.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"It was motivated by the observation that the (Yarowsky 95) algorithm added a very large number of rules in the first few iterations.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"\u2022 Similarly, when the na\u00efve Bayes classifier is configured such that the constituents require estimated probabilities strictly larger than 0.5 to be accepted, there is not enough probability mass remaining on crossing brackets for them to be included in the hypothesis.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"In our implementation, we make perhaps the simplest choice of weak hypothesis.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Approaches differ in the algorithms used for scoring and selecting the best path, as well as in the amount of contextual information used in the scoring process.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"\u2022 We evaluated translation from English, in addition to into English.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Gabbard and Kulick (2008) show that there is significant attachment ambiguity associated with iDafa, which occurs in 84.3% of the trees in our development set.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Given a set of n sentences, we can compute the sample mean x\ufffd and sample variance s2 of the individual sentence judgements xi: The extend of the confidence interval [x\u2212d, x+df can be computed by d = 1.96 \u00b7\ufffdn (6) Pairwise Comparison: As for the automatic evaluation metric, we want to be able to rank different systems against each other, for which we need assessments of statistical significance on the differences between a pair of systems.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"We present two algorithms.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"In total, across all domains, we kept 13,976 phrases with keywords.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"In the pinyin transliterations a dash(-) separates syllables that may be considered part of the same phonological word; spaces are used to separate plausible phonological words; and a plus sign (+) is used, where relevant, to indicate morpheme boundaries of interest.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"For example, we may have semantic information (e.g. database query operations) associated with the productions in a grammar.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"The first thing to note is that projectivizing helps in itself, even if no encoding is used, as seen from the fact that the projective baseline outperforms the non-projective training condition by more than half a percentage point on attachment score, although the gain is much smaller with respect to exact match.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"However, reads in the TRIE data structure are more expensive due to bit-level packing, so we found that it is faster to use interpolation search the entire time.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"Each pattern represents the role that a noun phrase plays in the surrounding context.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"We performed three experiments to evaluate our techniques.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"If the expression is a word or a short phrase (like \u00e2\u0080\u009ccorporation\u00e2\u0080\u009d and \u00e2\u0080\u009ccompany\u00e2\u0080\u009d), it is called a \u00e2\u0080\u009csynonym\u00e2\u0080\u009d.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"The combining algorithm is presented with the candidate parses and asked to choose which one is best.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"The manual scores are averages over the raw unnormalized scores.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"25 16.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"Pseudo-code describing the generalized boosting algorithm of Schapire and Singer is given in Figure 1.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"So, there is a limitation that IE can only be performed for a predefined task, like \u00e2\u0080\u009ccorporate mergers\u00e2\u0080\u009d or \u00e2\u0080\u009cmanagement succession\u00e2\u0080\u009d.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"To approximate these baselines, we implemented a very simple sentence selection algorithm in which parallel sentence pairs from OUT are ranked by the perplexity of their target half according to the IN language model.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"This essentially means that a better grammar tunes the joint model for optimized syntactic disambiguation at least in as much as their hyper parameters do.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"html 4 www.wagsoft.com\/RSTTool assigning rhetorical relations is a process loaded with ambiguity and, possibly, subjectivity.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"Such a classification can be seen as a not-always-correct summary of global features.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"However, it is possible to personify any noun, so in children's stories or fables, i\u00c2\u00a5JJ1l.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"attaching to terms denoting human beings.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"To conserve memory at the expense of accuracy, values may be quantized using q bits per probability and r bits per backoff6.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"The Expectation Maximization (EM) algorithm (Dempster, Laird and Rubin 77) is a common approach for unsupervised training; in this section we describe its application to the named entity problem.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"To make the projection practical, we rely on the twelve universal part-of-speech tags of Petrov et al. (2011).","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"(1992).","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"11 www.ling.unipotsdam.de\/sfb\/projekt a3.php 12 This step was carried out in the course of the diploma thesis work of David Reitter (2003), which de serves special mention here.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"For example, if Barry is not in commonWords and is found in the list of person first names, then the feature PersonFirstName will be set to 1.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"By this assumption, each element x E X can also be represented as (xi, x2) E X1 x X2.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"The space of a configuration is the sum of the lengths of the nonblank tape contents of the k work tapes.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"(a) 1 \u00c2\u00a7 . ;m t 7 leO z h e 4 pil m a 3 lu 4 sh an g4 bi ng 4 t h i s CL (assi fier) horse w ay on sic k A SP (ec t) 'This horse got sick on the way' (b) 1\u00c2\u00a7: . til y zhe4 tiao2 ma3lu4 hen3 shao3 this CL road very few 'Very few cars pass by this road' :$ chel jinglguo4 car pass by 2.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"For each pair we also record the context, i.e. the phrase between the two NEs (Step1).","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"needs to be in initCaps to be considered for this feature.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"This method, one instance of which we term the \"greedy algorithm\" in our evaluation of our own system in Section 5, involves starting at the beginning (or end) of the sentence, finding the longest word starting (ending) at that point, and then repeating the process starting at the next (previous) hanzi until the end (begin\u00c2\u00ad ning) of the sentence is reached.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Given n-gram counts {cn}Nn=1, we use Flog2 c1] bits per vocabulary identifier and Flog2 cn] per index into the table of ngrams.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Memory usage is the same as with binary search and lower than with set.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"Our model outperforms theirs on four out of five languages on the best hyperparameter setting and three out of five on the median setting, yielding an average absolute difference across languages of 12.9% and 3.9% for best and median settings respectively compared to their best EM or LBFGS performance.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"The sign test checks, how likely a sample of better and worse BLEU scores would have been generated by two systems of equal performance.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"Parser 3, the most accurate parser, was chosen 71% of the time, and Parser 1, the least accurate parser was chosen 16% of the time.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"Confidence Interval: Since BLEU scores are not computed on the sentence level, traditional methods to compute statistical significance and confidence intervals do not apply.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"pronunciations of individual words; they also need to compute intonational phrase boundaries in long utterances and assign relative prominence to words in those utterances.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"The development of automatic scoring methods is an open field of research.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"We dropped, however, one of the languages, Finnish, partly to keep the number of tracks manageable, partly because we assumed that it would be hard to find enough Finnish speakers for the manual evaluation.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"Similarly Figures 1 and 2 show how the isolated constituent precision varies by sentence length and the size of the span of the hypothesized constituent.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"The original tag set for the CoNLL-X Dutch data set consists of compounded tags that are used to tag multi-word units (MWUs) resulting in a tag set of over 300 tags.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"In our work, we demonstrate that using a simple na\u00c2\u00a8\u00c4\u00b1veBayes approach also yields substantial performance gains, without the associated training complexity.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"Another way to view the judgements is that they are less quality judgements of machine translation systems per se, but rankings of machine translation systems.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"Before explaining our method in detail, we present a brief overview in this subsection.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"This is the first time that we organized a large-scale manual evaluation.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"The current work treats both segmental and super-segmental phenomena, yet we note that there may be more adequate ways to treat supersegmental phenomena assuming Word-Based morphology as we explore in (Tsarfaty and Goldberg, 2008).","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"However, it is also mostly political content (even if not focused on the internal workings of the European Union) and opinion.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"The judgements tend to be done more in form of a ranking of the different systems.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"The sentences in the corpus were tagged by a transformation-based chunker and an NE tagger.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"For example, take ..., says Maury Cooper, a vice president at S.&P.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Table 2 Similarity matrix for segmentation judgments.","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"This is true of the widely used link grammar parser for English (Sleator and Temperley, 1993), which uses a dependency grammar of sorts, the probabilistic dependency parser of Eisner (1996), and more recently proposed deterministic dependency parsers (Yamada and Matsumoto, 2003; Nivre et al., 2004).","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"We further report SYNCS, the parsing metric of Cohen and Smith (2007), to facilitate the comparison.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Assuming unseen objects within each class are equiprobable, their probabilities are given by the Good-Turing theorem as: cis E( n'J.ls) Po oc N * E(N8ls) (2) where p815 is the probability of one unseen hanzi in class cls, E(n'J.15 ) is the expected number of hanzi in cls seen once, N is the total number of hanzi, and E(N(\/ 5 ) is the expected number of unseen hanzi in class cls.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"Using these states, we define partial hypothesis extensions, which are of the following type: (S0;C n fjg; j0) !","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"I = 1X21 N and N is a &quot;medium&quot; sized number so that it is feasible to collect 0(N) unlabeled examples.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"Next, we describe four contextual role knowledge sources that are created from the training examples and the caseframes.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"7).","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"This FSA I can be segmented into words by composing Id(I) with D*, to form the WFST shown in Figure 2(c), then selecting the best path through this WFST to produce the WFST in Figure 2(d).","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"When aligning the words in parallel texts (for language pairs like SpanishEnglish, French-English, ItalianGerman,...), we typically observe a strong localization effect.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"In our experiments, we used the same set of features as BergKirkpatrick et al. (2010): an indicator feature based In a traditional Markov model, the emission distribution P\u0398(Xi = xi  |Zi = zi) is a set of multinomials.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"We represent all morphological analyses of a given utterance using a lattice structure.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Time starts when Moses is launched and therefore includes model loading time.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"For this experiment, we compare our model with the uniform tag assignment prior (1TW) with the learned prior (+PRIOR).","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"3 These are not full case frames in the traditional sense, but they approximate a simple case frame with a single slot.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"If we were working with more than three parsers we could investigate minority constituents, those constituents that are suggested by at least one parser, but which the majority of the parsers do not suggest.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"i..f,..","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Motivated by these questions, we significantly raise baselines for three existing parsing models through better grammar engineering.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"To optimize this function, we used L-BFGS, a quasi-Newton method (Liu and Nocedal, 1989).","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"The returned state s(wn1) may then be used in a followon query p(wn+1js(wn1)) that extends the previous query by one word.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"We quantify error categories in both evaluation settings.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"Borth- wick (1999) successfully made use of other hand- coded systems as input for his MENE system, and achieved excellent results.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"As already pointed out in Section 2.4, current theories diverge not only on the number and definition of relations but also on apects of structure, i.e., whether a tree is sufficient as a representational device or general graphs are required (and if so, whether any restrictions can be placed on these graph\u00e2\u0080\u0099s structures \u00e2\u0080\u0094 cf.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"The NP is a complement to a preposition, which is the head of a PP.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"termined by the category of the word that follows it.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"That is, given a choice between segmenting a sequence abc into abc and ab, c, the former will always be picked so long as its cost does not exceed the summed costs of ab and c: while; it is possible for abc to be so costly as to preclude the larger grouping, this will certainly not usually be the case.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"We now describe the CoBoost algorithm for the named entity problem.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"The confidence intervals are computed by bootstrap resampling for BLEU, and by standard significance testing for the manual scores, as described earlier in the paper.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Different sentence structure and rich target language morphology are two reasons for this.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"As a partial solution, for pairs of hanzi that co-occur sufficiently often in our namelists, we use the estimated bigram cost, rather than the independence-based cost.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"We extend the Matsoukas et al approach in several ways.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"Before we turn to the evaluation, however, we need to introduce the data-driven dependency parser used in the latter experiments.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"In the named entity task, X1 might be the instance space for the spelling features, X2 might be the instance space for the contextual features.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"Evaluation results for links","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"The second stage links sets which involve the same pairs of individual NEs.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Since different judges judged different systems (recall that judges were excluded to judge system output from their own institution), we normalized the scores.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"The problem can be represented as a graph with 2N vertices corresponding to the members of X1 and X2.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"Our empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"Otherwise, label the training data with the combined spelling\/contextual decision list, then induce a final decision list from the labeled examples where all rules (regardless of strength) are added to the decision list.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"This is similar to using the Linux MAP POPULATE flag that is our default loading mechanism.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Thus in an English sentence such as I'm going to show up at the ACL one would reasonably conjecture that there are eight words separated by seven spaces.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"The Leaf Ancestor metric measures the cost of transforming guess trees to the reference (Sampson and Babarczy, 2003).","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"In fact, it is very difficult to maintain consistent standards, on what (say) an adequacy judgement of 3 means even for a specific language pair.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Eight out of the thirteen errors in the high frequency phrases in the CC-domain are the phrases in \u00e2\u0080\u009cagree\u00e2\u0080\u009d.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"Even with less training data, MENERGI outperforms Borthwick' s MENE + reference resolution (Borthwick, 1999).","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"12 One class of full personal names that this characterization does not cover are married women's names.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"In this shared task, we were also confronted with this problem, and since we had no funding for paying human judgements, we asked participants in the evaluation to share the burden.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"Ex: He was found in San Jose, where ...","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"To this end, we construct a bilingual graph over word types to establish a connection between the two languages (\u00a73), and then use graph label propagation to project syntactic information from English to the foreign language (\u00a74).","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"These are shown, with their associated costs, as follows: ABj nc 4.0 AB C\/jj 6.0 CD \/vb 5.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"For example, in MUC6, there are four zones (TXT, HL, DATELINE, DD).","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"suffixes (e.g., \u00ef\u00bf\u00bd=) Other notable parameters are second order vertical Markovization and marking of unary rules.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"We retain segmentation markers\u00e2\u0080\u0094which are consistent only in the vocalized section of the treebank\u00e2\u0080\u0094to differentiate between e.g. \u00ef\u00bf\u00bd \u00e2\u0080\u009cthey\u00e2\u0080\u009d and \u00ef\u00bf\u00bd + \u00e2\u0080\u009ctheir.\u00e2\u0080\u009d Because we use the vocalized section, we must remove null pronoun markers.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"Here we do not submit to this view.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"We asked participants to each judge 200\u2013300 sentences in terms of fluency and adequacy, the most commonly used manual evaluation metrics.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"The second algorithm extends ideas from boosting algorithms, designed for supervised learning tasks, to the framework suggested by (Blum and Mitchell 98).","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"We call such a constituent an isolated constituent.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Time for Moses itself to load, including loading the language model and phrase table, is included.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"The CoBoost algorithm described above divides the function Zco into two parts: Zco = 40 + 40.","label":0},{"source_text":"Here we present two algorithms.","target_text":"The problem is a binary classification problem.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"(2010) reports the best unsupervised results for English.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"In-domain Systran scores on this metric are lower than all statistical systems, even the ones that have much worse human scores.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"The performance was 80.99% recall and 61.83% precision.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"mark- ContainsVerb is especially effective for distinguishing root S nodes of equational sentences.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"So, this was a surprise element due to practical reasons, not malice.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Also surprising is the low test set OOV rate given the possibility of morphological variation in Arabic.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"First we mark any node that dominates (at any level) a verb sider POS tags when pre-terminals are the only intervening nodes between the nucleus and its bracketing (e.g., unaries, base NPs).","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"1,000 of these were picked at random, and labeled by hand to produce a test set.","label":0},{"source_text":"Here we present two algorithms.","target_text":"The input to AdaBoost is a set of training examples ((xi , yi), , (x\u201e.\u201e yrn)).","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"Initially, we planned to compare the semantic classes of an anaphor and a candidate and infer that they might be coreferent if their semantic classes intersected.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"We show that noun-noun vs. discourse-level coordination ambiguity in Arabic is a significant source of parsing errors (Table 8c).","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"For example, in Figure 3, we can see that the phrases in the \u00e2\u0080\u009cbuy\u00e2\u0080\u009d, \u00e2\u0080\u009cacquire\u00e2\u0080\u009d and \u00e2\u0080\u009cpurchase\u00e2\u0080\u009d sets are mostly paraphrases.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"The 95% confidence intervals for type-level errors are (5580, 9440) for the ATB and (1400, 4610) for the WSJ.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"The goal of machine translation is the translation of a text given in some source language into a target language.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Collisions between two keys in the table can be identified at model building time.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"2 60.7 56.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"Its only purpose is 3 This follows since each \u00ce\u00b8t has St \u00e2\u0088\u0092 1 parameters and.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"The probability of the classes assigned to the words in a sentence in a document is defined as follows: where is determined by the maximum entropy classifier.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"On the English side, however, the vertices (denoted by Ve) correspond to word types.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"basically complete, yet some improvements and extensions are still under way.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"Zt can be written as follows Following the derivation of Schapire and Singer, providing that W+ > W_, Equ.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"Recent work by Finkel and Manning (2009) which re-casts Daum\u00b4e\u2019s approach in a hierarchical MAP framework may be applicable to this problem.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":", for A. T.&T. nonalpha.. .","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"The above auxiliary quantity satisfies the following recursive DP equation: Qe0 (e; C; j) = Initial Skip Verb Final 1.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"We received submissions from 14 groups from 11 institutions, as listed in Figure 2.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"The domain is general politics, economics and science.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"We also thank ChaoHuang Chang, reviewers for the 1994 ACL conference, and four anonymous reviewers for Computational Linguistics for useful comments.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"In order to create an IE system for a new domain, one has to spend a long time to create the knowledge.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"18 77.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"The method being described-henceforth ST..","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"If two nouns have mutually exclusive semantic classes, then they cannot be coreferent.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Turning now to (1), we have the similar problem that splitting.into.ma3 'horse' andlu4 'way' is more costly than retaining this as one word .ma3lu4 'road.'","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"This PCFG is incorporated into the Stanford Parser, a factored model that chooses a 1-best parse from the product of constituency and dependency parses.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"Table 4 shows translation results for the three approaches.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"With the same Corporate- Suffix-List and Person-Prefix-List used in local features, for a token seen elsewhere in the same document with one of these suffixes (or prefixes), another feature Other-CS (or Other-PP) is set to 1.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"A name already mentioned previously in a document may appear in abbreviated form when it is mentioned again later.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"Furthermore, we do not connect the English vertices to each other, but only to foreign language vertices.4 The graph vertices are extracted from the different sides of a parallel corpus (De, Df) and an additional unlabeled monolingual foreign corpus Ff, which will be used later for training.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Two measures that can be used to compare judgments are: 1.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"Each decision determines the inclusion or exclusion of a candidate constituent.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"The types of patterns produced by AutoSlog are outlined in (Riloff, 1996).","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"While building a machine translation system is a serious undertaking, in future we hope to attract more newcomers to the field by keeping the barrier of entry as low as possible.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Quantization can be improved by jointly encoding probability and backoff.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"The table makes clear that enriching our grammar improves the syntactic performance as well as morphological disambiguation (segmentation and POS tagging) accuracy.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"It was also proposed to allow annotators to skip sentences that they are unable to judge.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"The hypothesized parse is then the set of constituents that are likely (P > 0.5) to be in the parse according to this model.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"8 66.4 52.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"19 We note that it is not always clear in Wang, Li, and Chang's examples which segmented words.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"First, we identify sources of syntactic ambiguity understudied in the existing parsing literature.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"This section describes AdaBoost, which is the basis for the CoBoost algorithm.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"How do additional ambiguities caused by devocalization affect statistical learning?","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"An extension of the TAG system was introduced by Joshi et al. (1975) and later redefined by Joshi (1987) in which the adjunction operation is defined on sets of elementary trees rather than single trees.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"The 3rd block contains the mixture baselines.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"We asked participants to each judge 200\u2013300 sentences in terms of fluency and adequacy, the most commonly used manual evaluation metrics.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"The parallel data came from the Europarl corpus (Koehn, 2005) and the ODS United Nations dataset (UN, 2006).","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"able at http:\/\/nlp.stanford.edu\/projects\/arabic.shtml.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"0 68.4 48.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"(We would like to note though that unlike previous boosting algorithms, the CoBoost algorithm presented here is not a boosting algorithm under Valiant's (Valiant 84) Probably Approximately Correct (PAC) model.)","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"If one of these checks fails then this knowledge source reports that the candidate is not a viable antecedent for the anaphor.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"Not every annotator was fluent in both the source and the target language.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"In Chinese text, individual characters of the script, to which we shall refer by their traditional name of hanzi,Z are written one after another with no intervening spaces; a Chinese sentence is shown in Figure 1.3 Partly as a result of this, the notion \"word\" has never played a role in Chinese philological tradition, and the idea that Chinese lacks any\u00c2\u00ad thing analogous to words in European languages has been prevalent among Western sinologists; see DeFrancis (1984).","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"All the NE pair instances which co-occur separated by at most 4 chunks are collected along with information about their NE types and the phrase between the NEs (the \u00e2\u0080\u0098context\u00e2\u0080\u0099).","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"In-domain Systran scores on this metric are lower than all statistical systems, even the ones that have much worse human scores.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"Lexicon and OOV Handling Our data-driven morphological-analyzer proposes analyses for unknown tokens as described in Section 5.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Each pattern represents the role that a noun phrase plays in the surrounding context.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"By taking the ratio of matching n-grams to the total number of n-grams in the system output, we obtain the precision pn for each n-gram order n. These values for n-gram precision are combined into a BLEU score: The formula for the BLEU metric also includes a brevity penalty for too short output, which is based on the total number of words in the system output c and in the reference r. BLEU is sensitive to tokenization.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"If the expression is a word or a short phrase (like \u00e2\u0080\u009ccorporation\u00e2\u0080\u009d and \u00e2\u0080\u009ccompany\u00e2\u0080\u009d), it is called a \u00e2\u0080\u009csynonym\u00e2\u0080\u009d.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"If one system is better in 95% of the sample sets, we conclude that its higher BLEU score is statistically significantly better.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"2 60.7 56.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"hanzi in the various name positions, derived from a million names.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"These methods demonstrated the benefits of incorporating linguistic features using a log-linear parameterization, but requires elaborate machinery for training.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"able at http:\/\/nlp.stanford.edu\/projects\/arabic.shtml.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"Sentences and systems were randomly selected and randomly shuffled for presentation.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"95 76.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"This is by now a fairly standard representation for multiple morphological segmentation of Hebrew utterances (Adler, 2001; Bar-Haim et al., 2005; Smith et al., 2005; Cohen and Smith, 2007; Adler, 2007).","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"The knowledge base then can be tested for its relation-inference capabilities on the basis of full-blown co-reference information.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"We substantially outperform all of them on query speed and offer lower memory consumption than lossless alternatives.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"2 70.7 52.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"This research was supported in part by the Defense Advanced Research Projects Agency as part of the Translingual Information Detection, Extraction and Summarization (TIDES) program, under Grant N66001001-18917 from the Space and Naval Warfare Systems Center, San Diego, and by the National Science Foundation under Grant IIS00325657.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"First, we identify sources of syntactic ambiguity understudied in the existing parsing literature.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"The toplevel weights are trained to maximize a metric such as BLEU on a small development set of approximately 1000 sentence pairs.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"In considering this aspect of a formalism, we hope to better understand the relationship between the structural descriptions generated by the grammars of a formalism, and the properties of semilinearity and polynomial recognizability.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"Such a classification can be seen as a not-always-correct summary of global features.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Pairwise comparison is done using the sign test.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"In the small window on the left, search queries can be entered, here one for an NP that has been annotated on the co-reference layer as bridging.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"The problem is to store these two values for a large and sparse set of n-grams in a way that makes queries efficient.","label":0},{"source_text":"There is no global pruning.","target_text":"A simple extension will be used to handle this problem.","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"Maximizing (7) is thus much faster than a typical MERT run. where co(s, t) are the counts from OUT, as in (6).","label":0},{"source_text":"There is no global pruning.","target_text":"3.2 Reordering with IBM Style.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Interpolation search has a more expensive pivot but performs less pivoting and reads, so it is slow on small data and faster on large data.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"We have shown the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"We develop our POS induction model based on the feature-based HMM of Berg-Kirkpatrick et al. (2010).","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"With some minor API changes, namely returning the length of the n-gram matched, it could also be faster\u2014though this would be at the expense of an optimization we explain in Section 4.1.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"While we have minimized forward-looking state in Section 4.1, machine translation systems could also benefit by minimizing backward-looking state.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Due to many similarly performing systems, we are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","label":1},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"This is done using a simple PCFG which is lexemebased.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Initially, we planned to compare the semantic classes of an anaphor and a candidate and infer that they might be coreferent if their semantic classes intersected.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"A different notion of information structure, is used in work such as that of (?), who tried to characterize felicitous constituent ordering (theme choice, in particular) that leads to texts presenting information in a natural, \u00e2\u0080\u009cflowing\u00e2\u0080\u009d way rather than with abrupt shifts of attention.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"'Malaysia.'","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"We therefore also normalized judgements on a per-sentence basis.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"This feature imposes constraints Table 1: Features based on the token string that are based on the probability of each name class during training.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"A number of PCC commentaries will be read by professional news speakers and prosodic features be annotated, so that the various annotation layers can be set into correspondence with intonation patterns.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"Finally, we intend to explore more sophisticated instanceweighting features for capturing the degree of generality of phrase pairs.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"The effect of a second reference resolution classifier is not entirely the same as that of global features.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"The translation scores for the hypotheses generated with different threshold values t0 are compared to the translation scores obtained with a conservatively large threshold t0 = 10:0 . For each test series, we count the number of sentences whose score is worse than the corresponding score of the test series with the conservatively large threshold t0 = 10:0, and this number is reported as the number of search errors.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"besuchen 9.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"The CoBoost algorithm described above divides the function Zco into two parts: Zco = 40 + 40.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"D o m ai n # of ph ras es t o t a l p h r a s e s ac cu ra cy C C 7 o r m o r e 1 0 5 8 7 . 6 % 6 o r l e s s 1 0 6 6 7 . 0 % P C 7 o r m o r e 3 5 9 9 9 . 2 % 6 o r l e s s 2 5 5 6 5 . 1 % Table 1.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"The present proposal falls into the last group.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"The final model tions.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Our coreference resolver also incorporates an existential noun phrase recognizer and a DempsterShafer probabilistic model to make resolution decisions.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"The belief value that would have been assigned to the intersection of these sets is .60*.70=.42, but this belief has nowhere to go because the null set is not permissible in the model.7 So this probability mass (.42) has to be redistributed.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Ends with the feminine affix :: p. 4.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"If the token is the first word of a sentence, then this feature is set to 1.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"The PROBING model can perform optimistic searches by jumping to any n-gram without needing state and without any additional memory.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"A graph D = (W, A) is well-formed iff it is acyclic and connected.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"Systems that generally do worse than others will receive a negative one.","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"It is worth noting that, although nonprojective constructions are less frequent in DDT than in PDT, they seem to be more deeply nested, since only about 80% can be projectivized with a single lift, while almost 95% of the non-projective arcs in PDT only require a single lift.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Moses keeps language models and many other resources in static variables, so these are still resident in memory.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"BABAR uses a named entity recognizer to identify proper names that refer to people and companies.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"6) are noisy, the results confirm that label propagation within the foreign language part of the graph adds significant quality for every language.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Beyond optimizing the memory size of TRIE, there are alternative data structures such as those in Guthrie and Hepple (2010).","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"An analysis of nouns that occur in both the singular and the plural in our database reveals that there is indeed a slight but significant positive correlation-R2 = 0.20, p < 0.005; see Figure 6.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Particular relations are also consistent with particular hypotheses about the segmentation of a given sentence, and the scores for particular relations can be incremented or decremented depending upon whether the segmentations with which they are consistent are \"popular\" or not.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"The first setting uses the European Medicines Agency (EMEA) corpus (Tiedemann, 2009) as IN, and the Europarl (EP) corpus (www.statmt.org\/europarl) as OUT, for English\/French translation in both directions.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"The second experiment is limited to data from PDT.5 The training part of the treebank was projectivized under different encoding schemes and used to train memory-based dependency parsers, which were run on the test part of the treebank, consisting of 7,507 sentences and 125,713 tokens.6 The inverse transformation was applied to the output of the parsers and the result compared to the gold standard test set.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"But diacritics are not present in unvocalized text, which is the standard form of, e.g., news media documents.3 VBD she added VP PUNC S VP VBP NP ...","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"The returned state s(wn1) may then be used in a followon query p(wn+1js(wn1)) that extends the previous query by one word.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"gaolxing4 'happy'=> F.i'JF.i'JJI!JI!","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"The computing time is given in terms of CPU time per sentence (on a 450MHz PentiumIIIPC).","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"This essentially means that a better grammar tunes the joint model for optimized syntactic disambiguation at least in as much as their hyper parameters do.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Using RandLM and the documented settings (8-bit values and 1 256 false-positive probability), we built a stupid backoff model on the same data as in Section 5.2.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"A step of an ATM consists of reading a symbol from each tape and optionally moving each head to the left or right one tape cell.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"See Section 5.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"A different knowledge source, called CFSemCFSem, compares the semantic expectations of the caseframe that extracts the anaphor with the semantic expectations of the caseframe that extracts the candidate.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"However, it is also mostly political content (even if not focused on the internal workings of the European Union) and opinion.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","label":1},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"We evaluate our approach on seven languages: English, Danish, Dutch, German, Portuguese, Spanish, and Swedish.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"Initially, we planned to compare the semantic classes of an anaphor and a candidate and infer that they might be coreferent if their semantic classes intersected.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"For example, suppose the current model assigns a belief value of .60 to {A, B}, meaning that it is 60% sure that the correct hypothesis is either A or B. Then new evidence arrives with a belief value of .70 assigned 5 Initially there are no competing hypotheses because all hypotheses are included in \u00ce\u00b8 by definition.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"9 65.5 46.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"We consider the unsupervised POS induction problem without the use of a tagging dictionary.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"Richer tag sets have been suggested for modeling morphologically complex distinctions (Diab, 2007), but we find that linguistically rich tag sets do not help parsing.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"In figure 4, reverse relations are indicated by `*\u00e2\u0080\u0099 next to the frequency.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"In the named entity domain these rules were Each of these rules was given a strength of 0.9999.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"Table 5 provides insight into the behavior of different models in terms of the tagging lexicon they generate.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"Step 2.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"each word in the lexicon whether or not each string is actually an instance of the word in question.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"6 Conclusions.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"A geometrical progression of language families defined by Weir (1987) involves tree sets with increasingly complex path sets.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"2.5 Connectives with scopes.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"Across all languages, high performance can be attained by selecting a single tag per word type.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"The wounds are still healing.), entity-attribute (e.g., She 2001), who determined that in their corpus of German computer tests, 38% of relations were lexically signalled.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"Our method does not assume any knowledge about the target language (in particular no tagging dictionary is assumed), making it applicable to a wide array of resource-poor languages.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"buy - acquire (5) buy - agree (2) buy - purchase (5) buy - acquisition (7) buy - pay (2)* buy - buyout (3) buy - bid (2) acquire - purchase (2) acquire - acquisition (2) acquire - pay (2)* purchase - acquisition (4) purchase - stake (2)* acquisition - stake (2)* unit - subsidiary (2) unit - parent (5) It is clear that these links form two clusters which are mostly correct.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"A corpus of German newspaper commentaries has been assembled and annotated with different information (and currently, to different degrees): part-of-speech, syntax, rhetorical structure, connectives, co-reference, and information structure.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"For each trigram type x2 x3 x4 in a sequence x1 x2 x3 x4 x5, we count how many times that trigram type co-occurs with the different instantiations of each concept, and compute the point-wise mutual information (PMI) between the two.5 The similarity between two trigram types is given by summing over the PMI values over feature instantiations that they have in common.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"Lemma: If the number of votes required by constituent voting is greater than half of the parsers under consideration the resulting structure has no crossing constituents.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"Such tag sequences are often treated as \u201ccomplex tags\u201d (e.g.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Nonetheless, the results of the comparison with human judges demonstrates that there is mileage being gained by incorporating models of these types of words.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"(1999) did not report using any dictionaries, but mentioned in a footnote that they have added list membership features, which have helped marginally in certain domains.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"Not every annotator was fluent in both the source and the target language.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"This fully generative model caters for real interaction between the syntactic and morphological levels as a part of a single coherent process.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"For.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"4 To be sure, it is not always true that a hanzi represents a syllable or that it represents a morpheme.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"Pseudo-code describing the generalized boosting algorithm of Schapire and Singer is given in Figure 1.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"f, nan2gual+men0 'pumpkins' is by no means impossible.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"More judgements would have enabled us to make better distinctions, but it is not clear what the upper limit is.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"To show that the derivation tree set of a TAG is a local set, nodes are labeled by pairs consisting of the name of an elementary tree and the address at which it was adjoined, instead of labelling edges with addresses.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"AdaBoost finds a weighted combination of simple (weak) classifiers, where the weights are chosen to minimize a function that bounds the classification error on a set of training examples.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"2.2 Syntactic structure.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"That is, given a choice between segmenting a sequence abc into abc and ab, c, the former will always be picked so long as its cost does not exceed the summed costs of ab and c: while; it is possible for abc to be so costly as to preclude the larger grouping, this will certainly not usually be the case.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"It is possible one could produce better models by introducing features describing constituents and their contexts because one parser could be much better than the majority of the others in particular situations.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"This formulation of the constraint feature is equivalent to the use of a tagging dictionary extracted from the graph using a threshold T on the posterior distribution of tags for a given word type (Eq.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"This is especially true in the case of quotations\u00e2\u0080\u0094which are common in the ATB\u00e2\u0080\u0094where (1) will follow a verb like (2) (Figure 1).","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Since the segmentation corresponds to the sequence of words that has the lowest summed unigram cost, the segmenter under discussion here is a zeroth-order model.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"In all figures, we present the per-sentence normalized judgements.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"This latter evaluation compares the performance of the system with that of several human judges since, as we shall show, even people do not agree on a single correct way to segment a text.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"(Other classes handled by the current system are discussed in Section 5.)","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"The sentence-selection approach is crude in that it imposes a binary distinction between useful and non-useful parts of OUT.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"In Input: {(x1,i, Initialize: Vi, j : e(xi) = 0.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"The set of n-grams appearing in a model is sparse, and we want to efficiently find their associated probabilities and backoff penalties.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"However, it is possible to personify any noun, so in children's stories or fables, i\u00c2\u00a5JJ1l.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"(In this figure eps is c) be implemented, though, such as a maximal-grouping strategy (as suggested by one reviewer of this paper); or a pairwise-grouping strategy, whereby long sequences of unattached hanzi are grouped into two-hanzi words (which may have some prosodic motivation).","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"Although embedding this version of LCFRS's in the framework of ILFP developed by Rounds (1985) is straightforward, our motivation was to capture properties shared by a family of grammatical systems and generalize them defining a class of related formalisms.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Analysis of the data revealed that the contextual role knowledge is especially helpful for resolving pronouns because, in general, they are semantically weaker than definite NPs.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"Different annotations of the same text are mapped into the same data structure, so that search queries can be formulated across annotation levels.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"ni2ya3 and @5:2 xilya3, respectively.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"The range of the score is between 0 and 1 (higher is better).","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"To be short, we omit the target words e; e0 in the formulation of the search hypotheses.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"Unfortunately, Yarowsky's method is not well understood from a theoretical viewpoint: we would like to formalize the notion of redundancy in unlabeled data, and set up the learning task as optimization of some appropriate objective function.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"Output of the learning algorithm: a function h:Xxy [0, 1] where h(x, y) is an estimate of the conditional probability p(y1x) of seeing label y given that feature x is present.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"Presenting the output of several system allows the human judge to make more informed judgements, contrasting the quality of the different systems.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"The government has to make a decision, and do it quickly.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Linear probing hash tables must have more buckets than entries, or else an empty bucket will never be found.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"Now we have sets of phrases which share a keyword and we have links between those sets.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"By design, they readily capture regularities at the token-level.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"As is standard, we use a fixed constant K for the number of tagging states.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"Using less training data than other systems, our NER is able to perform as well as other state-of-the-art NERs.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"In many cases, inspection of either the spelling or context alone is sufficient to classify an example.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"(S2) The burglar fired the gun three times and fled.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"(See Sproat and Shih 1995.)","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"We offer a state function s(wn1) = wn\ufffd where substring wn\ufffd is guaranteed to extend (to the right) in the same way that wn1 does for purposes of language modeling.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"Our assumption is that caseframes that co-occur in resolutions often have a 2 This normalization is performed syntactically without semantics, so the agent and patient roles are not guaranteed to hold, but they usually do in practice.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"Specifically, for the ith word type, the set of token-level tags associated with token occurrences of this word, denoted t(i), must all take the value Ti to have nonzero mass. Thus in the context of Gibbs sampling, if we want to block sample Ti with t(i), we only need sample values for Ti and consider this setting of t(i).","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"If the semantic expectations do not intersect, then we know that the case- frames extract mutually exclusive types of noun phrases.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Because b is a function, no additional hypothesis splitting happens.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"There is a guarantee of no crossing brackets but there is no guarantee that a constituent in the tree has the same children as it had in any of the three original parses.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"The learned patterns are then normalized and applied to the corpus.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"This decreases the statistical significance of our results compared to those studies.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"Figure 2 shows examples of extracted NE pair instances and their contexts.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"Word Head Of Complement POS 1 '01 inna \u00e2\u0080\u009cIndeed, truly\u00e2\u0080\u009d VP Noun VBP 2 '01 anna \u00e2\u0080\u009cThat\u00e2\u0080\u009d SBAR Noun IN 3 01 in \u00e2\u0080\u009cIf\u00e2\u0080\u009d SBAR Verb IN 4 01 an \u00e2\u0080\u009cto\u00e2\u0080\u009d SBAR Verb IN Table 1: Diacritized particles and pseudo-verbs that, after orthographic normalization, have the equivalent surface form 0 an.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"The importance of dictionaries in NERs has been investigated in the literature (Mikheev et al., 1999).","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"We aligned the texts at a sentence level across all four languages, resulting in 1064 sentence per language.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Language models are widely applied in natural language processing, and applications such as machine translation make very frequent queries.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"However, there is again local grammatical information that should favor the split in the case of (1a): both .ma3 'horse' and .ma3 lu4 are nouns, but only .ma3 is consistent with the classifier pil, the classifier for horses.21 By a similar argument, the preference for not splitting , lm could be strengthened in (lb) by the observation that the classifier 1'1* tiao2 is consistent with long or winding objects like , lm ma3lu4 'road' but not with,ma3 'horse.'","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"Given a set of n sentences, we can compute the sample mean x\ufffd and sample variance s2 of the individual sentence judgements xi: The extend of the confidence interval [x\u2212d, x+df can be computed by d = 1.96 \u00b7\ufffdn (6) Pairwise Comparison: As for the automatic evaluation metric, we want to be able to rank different systems against each other, for which we need assessments of statistical significance on the differences between a pair of systems.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"2.1 Overview.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"The first two rows of the table are baselines.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"The general-language features have a slight advantage over the similarity features, and both are better than the SVM feature.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"The global feature groups are: InitCaps of Other Occurrences (ICOC): There are 2 features in this group, checking for whether the first occurrence of the same word in an unambiguous position (non first-words in the TXT or TEXT zones) in the same document is initCaps or not-initCaps.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"Relative pronouns with only 1 NP in scope..","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"For humans, this characteristic can impede the acquisition of literacy.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"(b) F.i'JJI!","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"We use a solution to this problem similar to the one presented in (Och et al., 1999), where target words are joined during training.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"Our full model outperforms the \u201cNo LP\u201d setting because it has better vocabulary coverage and allows the extraction of a larger set of constraint features.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"Acknowledgments We thank Meni Adler and Michael Elhadad (BGU) for helpful comments and discussion.","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"We will briefly discuss this point in Section 3.1.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"Fall.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"One implementation issue deserves some elaboration.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Other kinds of productive word classes, such as company names, abbreviations (termed fijsuolxie3 in Mandarin), and place names can easily be 20 Note that 7 in E 7 is normally pronounced as leO, but as part of a resultative it is liao3..","label":1},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Generalizing state minimization, the model could also provide explicit bounds on probability for both backward and forward extension.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"We will refer to the semantic classes that co-occur with a caseframe as the semantic expectations of the caseframe.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"We can check, what the consequences of less manual annotation of results would have been: With half the number of manual judgements, we can distinguish about 40% of the systems, 10% less.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"If the parse contains productions from outside our grammar the machine has no direct method for handling them (e.g. the resulting database query may be syntactically malformed).","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"Second, the reduced number of hidden variables and parameters dramatically speeds up learning and inference.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"To eliminate such sequences, we define a transition probability between word classes to be equal to 1 if the sequence is admissible, and 0 otherwise.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"The second algorithm builds on a boosting algorithm called AdaBoost (Freund and Schapire 97; Schapire and Singer 98).","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"Pairwise comparison is done using the sign test.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"In words, the judgements are normalized, so that the average normalized judgement per judge is 3.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Compared with the widely- SRILM, our is 2.4 times as fast while using 57% of the mem- The structure is a trie with bit-level packing, sorted records, interpolation search, and optional quantization aimed lower memory consumption. simultaneously uses less memory than the smallest lossless baseline and less CPU than the baseline.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"C is the union of the sets of constituents suggested by the parsers. r(c) is a binary function returning t (for true) precisely when the constituent c E C should be included in the hypothesis.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"Given an anaphor, BABAR identifies the caseframe that would extract it from its sentence.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"To investigate the influence of these factors, we analyze Modern Standard Arabic (henceforth MSA, or simply \u00e2\u0080\u009cArabic\u00e2\u0080\u009d) because of the unusual opportunity it presents for comparison to English parsing results.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"Cluster phrases based on Links We now have a set of phrases which share a keyword.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"Our full model outperforms the \u201cNo LP\u201d setting because it has better vocabulary coverage and allows the extraction of a larger set of constraint features.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"The distinctions in the ATB are linguistically justified, but complicate parsing.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"Figure 1: Translation of PCC sample commentary (STTS)2.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"This larger corpus was kindly provided to us by United Informatics Inc., R.O.C. a set of initial estimates of the word frequencies.9 In this re-estimation procedure only the entries in the base dictionary were used: in other words, derived words not in the base dictionary and personal and foreign names were not used.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Figure 5 shows how this model is implemented as part of the dictionary WFST.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"Arabic sentences of up to length 63 would need to be.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"f, nan2gual+men0 'pumpkins' is by no means impossible.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"collected too.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"For each label (Per s on, organization and Location), take the n contextual rules with the highest value of Count' (x) whose unsmoothed3 strength is above some threshold pmin.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"The following algorithm was then used to induce new rules: Let Count' (x) be the number of times feature x is seen with some known label in the training data.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Figure 1 provides some statistics about this corpus.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"As is described in Section 4, the property of having a derivation tree set that is a local set appears to be useful in showing important properties of the languages generated by the formalisms.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"For the translation model Pr(fJ 1 jeI 1), we go on the assumption that each source word is aligned to exactly one target word.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"This is the form of recursive levels in iDafa constructs.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"(Yarowsky 95) describes an algorithm for word-sense disambiguation that exploits redundancy in contextual features, and gives impressive performance.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Ex: The brigade, which attacked ...","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"However, we have reason to doubt Chang et al.'s performance claims.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"In the ATB, :: b asta\u00e2\u0080\u0099adah is tagged 48 times as a noun and 9 times as verbal noun.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"For developing these mechanisms, the possibility to feed in hand-annotated information is very useful.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"We use the log-linear tagger of Toutanova et al.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"3.2 The DempsterShafer Decision Model.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"The difference between the featureless model (+PRIOR) and our full model (+FEATS) is 13.6% and 7.7% average error reduction on best and median settings respectively.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"We present KenLM, a library that implements two data structures for efficient language model queries, reducing both time and costs.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"The development of a na\u00efve Bayes classifier involves learning how much each parser should be trusted for the decisions it makes.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Since the segmentation corresponds to the sequence of words that has the lowest summed unigram cost, the segmenter under discussion here is a zeroth-order model.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"We evaluate our approach on seven languages: English, Danish, Dutch, German, Portuguese, Spanish, and Swedish.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"We used 22 features for the logistic weighting model, divided into two groups: one intended to reflect the degree to which a phrase pair belongs to general language, and one intended to capture similarity to the IN domain.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"Our model outperforms theirs on four out of five languages on the best hyperparameter setting and three out of five on the median setting, yielding an average absolute difference across languages of 12.9% and 3.9% for best and median settings respectively compared to their best EM or LBFGS performance.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"It was motivated by the observation that the (Yarowsky 95) algorithm added a very large number of rules in the first few iterations.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"The way judgements are collected, human judges tend to use the scores to rank systems against each other.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"In the CC-domain, there are 32 sets of phrases which contain more than 2 phrases.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"When SRILM estimates a model, it sometimes removes n-grams but not n + 1-grams that extend it to the left.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"We first define &quot;pseudo-labels&quot;,-yt, as follows: = Yi t sign(g 0\\ 2\u2014 kx2,m < i < n Thus the first m labels are simply copied from the labeled examples, while the remaining (n \u2014 m) examples are taken as the current output of the second classifier.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"If they are found in a list, then a feature for that list will be set to 1.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"(Brin 98) ,describes a system for extracting (author, book-title) pairs from the World Wide Web using an approach that bootstraps from an initial seed set of examples.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"As each global feature group is added to the list of features, we see improvements to both MUC6 and MUC6 MUC7 Baseline 90.75% 85.22% + ICOC 91.50% 86.24% + CSPP 92.89% 86.96% + ACRO 93.04% 86.99% + SOIC 93.25% 87.22% + UNIQ 93.27% 87.24% Table 3: F-measure after successive addition of each global feature group Table 5: Comparison of results for MUC6 Systems MUC6 MUC7 No.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"This style of naming is never required and seems to be losing currency.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"The annotator can then \u00e2\u0080\u009cclick away\u00e2\u0080\u009d those words that are here not used as connectives (such as the conjunction und (\u00e2\u0080\u0098and\u00e2\u0080\u0099) used in lists, or many adverbials that are ambiguous between connective and discourse particle).","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Table 8a shows that the best model recovers SBAR at only 71.0% F1.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"An important reason for separating the two types of features is that this opens up the possibility of theoretical analysis of the use of unlabeled examples.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"We then extract a set of possible tags tx(y) by eliminating labels whose probability is below a threshold value \u03c4: We describe how we choose \u03c4 in \u00a76.4.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"When the maSdar lacks a determiner, the constituent as a whole resem bles the ubiquitous annexation construct \u00ef\u00bf\u00bd ?f iDafa.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Their default variant implements a forward trie, in which words are looked up in their natural left-to-right order.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"In this section, we try to compare our results with those obtained by IdentiFinder ' 97 (Bikel et al., 1997), IdentiFinder ' 99 (Bikel et al., 1999), and MENE (Borthwick, 1999).","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"We can do that . Input: Das ist zu knapp , weil ich ab dem dritten in Kaiserslautern bin . Genaugenommen nur am dritten . Wie ware es denn am ahm Samstag , dem zehnten Februar ? MonS: That is too tight , because I from the third in Kaiserslautern . In fact only on the third . How about ahm Saturday , the tenth of February ? QmS: That is too tight , because I am from the third in Kaiserslautern . In fact only on the third . Ahm how about Saturday , February the tenth ? IbmS: That is too tight , from the third because I will be in Kaiserslautern . In fact only on the third . Ahm how about Saturday , February the tenth ? Input: Wenn Sie dann noch den siebzehnten konnten , ware das toll , ja . MonS: If you then also the seventeenth could , would be the great , yes . QmS: If you could then also the seventeenth , that would be great , yes . IbmS: Then if you could even take seventeenth , that would be great , yes . Input: Ja , das kommt mir sehr gelegen . Machen wir es dann am besten so . MonS: Yes , that suits me perfectly . Do we should best like that . QmS: Yes , that suits me fine . We do it like that then best . IbmS: Yes , that suits me fine . We should best do it like that .","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"Like HG's, TAG's, and MCTAG's, members of LCFRS can manipulate structures more complex than terminal strings and use composition operations that are more complex that concatenation.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"Mixing, smoothing, and instance-feature weights are learned at the same time using an efficient maximum-likelihood procedure that relies on only a small in-domain development corpus.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"The resulting parsers surpass the best previously published performance results for the Penn Treebank.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"MUC7 has also seen hybrids of statistical NERs and hand-coded systems (Mikheev et al., 1998; Borthwick, 1999), notably Mikheev' s system, which achieved the best performance of 93.39% on the official NE test data.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"In our particular case the majority requires the agreement of only two parsers because we have only three.","label":0},{"source_text":"Here we present two algorithms.","target_text":"Pseudo-labels are formed by taking seed labels on the labeled examples, and the output of the fixed classifier on the unlabeled examples.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"Furthermore, the combination of pruning and vertical markovization of the grammar outperforms the Oracle results reported by Cohen and Smith.","label":0},{"source_text":"Here we present two algorithms.","target_text":"For the experiments in this paper we made a couple of additional modifications to the CoBoost algorithm.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"We are currently exploring such algorithms.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"In our forth model GTnph we add the definiteness status of constituents following Tsarfaty and Sima\u2019an (2007).","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"In Figure 4, we displayed the number of system comparisons, for which we concluded statistical significance.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"While possible to utilize the feature-based log-linear approach described in Berg-Kirkpatrick et al.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"In the pinyin transliterations a dash(-) separates syllables that may be considered part of the same phonological word; spaces are used to separate plausible phonological words; and a plus sign (+) is used, where relevant, to indicate morpheme boundaries of interest.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"Different sentence structure and rich target language morphology are two reasons for this.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"We believe it is natural for authors to use abbreviations in subsequent mentions of a named entity (i.e., first \u00e2\u0080\u009cPresident George Bush\u00e2\u0080\u009d then \u00e2\u0080\u009cBush\u00e2\u0080\u009d).","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"e0; e are the last two target words, C is a coverage set for the already covered source positions and j is the last position visited.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"+ cost(unseen(fm, as desired.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"The type of alignment we have considered so far requires the same length for source and target sentence, i.e. I = J. Evidently, this is an unrealistic assumption, therefore we extend the concept of inverted alignments as follows: When adding a new position to the coverage set C, we might generate either \u00c3\u0086 = 0 or \u00c3\u0086 = 1 new target words.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"The Input The set of analyses for a token is thus represented as a lattice in which every arc corresponds to a specific lexeme l, as shown in Figure 1.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"(a) Of the high frequency phrasal categories, ADJP and SBAR are the hardest to parse.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"(2010) consistently outperforms ours on English, we obtain substantial gains across other languages.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"Our second contribution is to apply instance weighting at the level of phrase pairs.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"For that application, at a minimum, one would want to know the phonological word boundaries.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"The weak learner for two-class problems computes a weak hypothesis h from the input space into the reals (h : 2x -4 R), where the sign4 of h(x) is interpreted as the predicted label and the magnitude I h(x)I is the confidence in the prediction: large numbers for I h(x)I indicate high confidence in the prediction, and numbers close to zero indicate low confidence.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Our test machine has two Intel Xeon E5410 processors totaling eight cores, 32 GB RAM, and four Seagate Barracuda disks in software RAID 0 running Linux 2.6.18.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"Moving beyond directly related work, major themes in SMT adaptation include the IR (Hildebrand et al., 2005; L\u00a8u et al., 2007; Zhao et al., 2004) and mixture (Finch and Sumita, 2008; Foster and Kuhn, 2007; Koehn and Schroeder, 2007; L\u00a8u et al., 2007) approaches for LMs and TMs described above, as well as methods for exploiting monolingual in-domain text, typically by translating it automatically and then performing self training (Bertoldi and Federico, 2009; Ueffing et al., 2007; Schwenk and Senellart, 2009).","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"We are also grateful to Markus Dickinson, Ali Farghaly, Nizar Habash, Seth Kulick, David McCloskey, Claude Reichard, Ryan Roth, and Reut Tsarfaty for constructive discussions.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Yet, some hanzi are far more probable in women's names than they are in men's names, and there is a similar list of male-oriented hanzi: mixing hanzi from these two lists is generally less likely than would be predicted by the independence model.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"Our results suggest that it is possible to learn accurate POS taggers for languages which do not have any annotated data, but have translations into a resource-rich language.","label":1},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"It is difficult to motivate these days why one ministry should be exempt from cutbacks \u00e2\u0080\u0094 at the expense of the others.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Table 2 Similarity matrix for segmentation judgments.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"Here, the term frequency (TF) is the frequency of a word in the bag and the inverse term frequency (ITF) is the inverse of the log of the frequency in the entire corpus.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"This paper describes the several performance techniques used and presents benchmarks against alternative implementations.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"For example, co-occurring caseframes may reflect synonymy (e.g., \u00e2\u0080\u009c<patient> kidnapped\u00e2\u0080\u009d and \u00e2\u0080\u009c<patient> abducted\u00e2\u0080\u009d) or related events (e.g., \u00e2\u0080\u009c<patient> kidnapped\u00e2\u0080\u009d and \u00e2\u0080\u009c<patient> released\u00e2\u0080\u009d).","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"The details are given in (Och and Ney, 2000).","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"Process nominals name the action of the transitive or ditransitive verb from which they derive.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"For the TM, this is: where cI(s, t) is the count in the IN phrase table of pair (s, t), po(s|t) is its probability under the OUT TM, and cI(t) = &quot;s, cI(s', t).","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"The Recency KS computes the distance between the candidate and the anaphor relative to its scope.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"The basic strategy is, for a given pair of entity types, to start with some examples, like several famous book title and author pairs; and find expressions which contains those names; then using the found expressions, find more author and book title pairs.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"The word joining is done on the basis of a likelihood criterion.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"For example Chen and Liu (1992) report precision and recall rates of over 99%, but this counts only the words that occur in the test corpus that also occur in their dictionary.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"Almost all annotators expressed their preference to move to a ranking-based evaluation in the future.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"The linear LM (lin lm), TM (lin tm) and MAP TM (map tm) used with non-adapted counterparts perform in all cases slightly worse than the log-linear combination, which adapts both LM and TM components.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"This is because in many cases, the use of hyphens can be considered to be optional (e.g., third-quarter or third quarter).","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"Every token is independent of the others, and the sentence lattice is in fact a concatenation of smaller lattices, one for each token.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"This feature includes named entities, which the ATB marks with a flat NP node dominating an arbitrary number of NNP pre-terminal daughters (Figure 2).","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"It was filtered to retain the top 30 translations for each source phrase using the TM part of the current log-linear model.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"We have tested the translation system on the Verbmobil task (Wahlster 1993).","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"The observed performance gains, coupled with the simplicity of model implementation, makes it a compelling alternative to existing more complex counterparts.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Further, the special hash 0 suffices to flag empty buckets.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"The types of patterns produced by AutoSlog are outlined in (Riloff, 1996).","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"The kind of annotation work presented here would clearly benefit from the emergence of standard formats and tag sets, which could lead to sharable resources of larger size.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"In our implementation, we make perhaps the simplest choice of weak hypothesis.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"It was motivated by the observation that the (Yarowsky 95) algorithm added a very large number of rules in the first few iterations.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"Since each zi is a contiguous substring of the input (say ai,), and no two substrings overlap, we can represent zi by the pair of integers (i2, i2).","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"(For some recent corpus-based work on Chinese abbreviations, see Huang, Ahrens, and Chen [1993].)","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"D o m ai n Li n k ac cu ra cy W N c o v e r a g e C C 7 3 . 3 % 2 \/ 1 1 P C 8 8 . 9 % 2 \/ 8 Table 2.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"The feature-based model replaces the emission distribution with a log-linear model, such that: on the word identity x, features checking whether x contains digits or hyphens, whether the first letter of x is upper case, and suffix features up to length 3.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"It has no syntactic function.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"If the expression is a word or a short phrase (like \u00e2\u0080\u009ccorporation\u00e2\u0080\u009d and \u00e2\u0080\u009ccompany\u00e2\u0080\u009d), it is called a \u00e2\u0080\u009csynonym\u00e2\u0080\u009d.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"How should the absence of vowels and syntactic markers influence annotation choices and grammar development?","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"123 examples fell into the noise category.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"HR0011-06-C-0022.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Let us consider an example of ambiguity caused by devocalization.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"This smooth guarantees that there are no zeroes estimated.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"On the other hand, when all systems produce muddled output, but one is better, and one is worse, but not completely wrong, a judge is inclined to hand out judgements of 4, 3, and 2.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"There is twice as much language modelling data, since training data for the machine translation system is filtered against sentences of length larger than 40 words.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"(2) was extended to have an additional, innermost loop over the (3) possible labels.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"If one of these checks fails then this knowledge source reports that the candidate is not a viable antecedent for the anaphor.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"While Berg-Kirkpatrick et al.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"Only tokens with initCaps not found in commonWords are tested against each list in Table 2.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"A cell in the bottom row of the parse chart is required for each potential whitespace boundary.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"This approach is compared to another reordering scheme presented in (Berger et al., 1996).","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"(2009b) evaluated the Bikel parser using the same ATB split, but only reported dev set results with gold POS tags for sentences of length \u00e2\u0089\u00a4 40.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Still, for about good number of sentences, we do have this direct comparison, which allows us to apply the sign test, as described in Section 2.2.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"Still, for about good number of sentences, we do have this direct comparison, which allows us to apply the sign test, as described in Section 2.2.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"Models To assess the marginal utility of each component of the model (see Section 3), we incremen- tally increase its sophistication.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Virginia) and -sia are normally transliterated as fbSi!","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"8 66.4 52.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"While our method also enforces a singe tag per word constraint, it leverages the transition distribution encoded in an HMM, thereby benefiting from a richer representation of context.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"It is important to bear in mind, though, that this is not an inherent limitation of the model.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"This can be seen as a rough approximation of Yarowsky and Ngai (2001).","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"In this case, Maury Cooper is extracted.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"Again, the idea is that having a picture of syntax, co-reference, and sentence-internal information structure at one\u00e2\u0080\u0099s disposal should aid in finding models of discourse structure that are more explanatory and can be empirically supported.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"We use graph-based label propagation for cross-lingual knowledge transfer and use the projected labels as features in an unsupervised model (Berg- Kirkpatrick et al., 2010).","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"For example, given a sequence F1G1G2, where F1 is a legal single-hanzi family name, and Plural Nouns X g 0 g \"' X X 0 T!i c\"'.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Minimal perfect hashing is used to find the index at which a quantized probability and possibly backoff are stored.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"4.1 Local Features.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"Otherwise, it is set to 0.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"We evaluate the time and memory consumption of each data structure by computing perplexity on 4 billion tokens from the English Gigaword corpus (Parker et al., 2009).","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"0 Figure 5 An example of affixation: the plural affix.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Let us notate the set of previously unseen, or novel, members of a category X as unseen(X); thus, novel members of the set of words derived in f, menO will be de\u00c2\u00ad noted unseen(f,).","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"Thus we are interested not in extraction, but actual generation from representations that may be developed to different degrees of granularity.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"Instead, we condition on the type-level tag assignments T . Specifically, let St = {i|Ti = t} denote the indices of theword types which have been assigned tag t accord ing to the tag assignments T . Then \u00ce\u00b8t is drawn from DIRICHLET(\u00ce\u00b1, St), a symmetric Dirichlet which only places mass on word types indicated by St. This ensures that each word will only be assigned a single tag at inference time (see Section 4).","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"Morphological disambiguators that consider a token in context (an utterance) and propose the most likely morphological analysis of an utterance (including segmentation) were presented by Bar-Haim et al. (2005), Adler and Elhadad (2006), Shacham and Wintner (2007), and achieved good results (the best segmentation result so far is around 98%).","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"Our model outperforms theirs on four out of five languages on the best hyperparameter setting and three out of five on the median setting, yielding an average absolute difference across languages of 12.9% and 3.9% for best and median settings respectively compared to their best EM or LBFGS performance.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"Our representation of contextual roles is based on information extraction patterns that are converted into simple caseframes.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"2.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"And then there are decisions that systems typically hard-wire, because the linguistic motivation for making them is not well understood yet.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"Similar behavior is observed when adding features.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"This research was supported in part by the Defense Advanced Research Projects Agency as part of the Translingual Information Detection, Extraction and Summarization (TIDES) program, under Grant N66001001-18917 from the Space and Naval Warfare Systems Center, San Diego, and by the National Science Foundation under Grant IIS00325657.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Mohri [1995]) shows promise for improving this situation.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"We use the default inference parameters.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"Ties are rare in Bayes switching because the models are fine-grained \u2014 many estimated probabilities are involved in each decision.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"If two systems\u2019 scores are close, this may simply be a random effect in the test data.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"The TRIE model continues to use the least memory of ing (-P) with MAP POPULATE, the default.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"P St = n. \u00ce\u00b2 T VARIABLES \u00cf\u0088 Y W : Word types (W1 ,.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"The contextual role knowledge had the greatest impact on pronouns: +13% recall for terrorism and +15% recall for disasters, with a +1% precision gain in terrorism and a small precision drop of -3% in disasters.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Still, for about good number of sentences, we do have this direct comparison, which allows us to apply the sign test, as described in Section 2.2.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"It can be seen that M performs a top-down recognition of the input al ... nin logspace.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"Finally, the concatenated 5 * 20% output is used to train the reference resolution component.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"When a company buys another company, a paying event can occur, but these two phrases do not indicate the same event.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Gan's solution depends upon a fairly sophisticated language model that attempts to find valid syntactic, semantic, and lexical relations between objects of various linguistic types (hanzi, words, phrases).","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"In the small window on the left, search queries can be entered, here one for an NP that has been annotated on the co-reference layer as bridging.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"This is the first set that gives us a fair evaluation of the Bayes models, and the Bayes switching model performs significantly better than its non-parametric counterpart.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"Two annotators received training with the RST definitions and started the process with a first set of 10 texts, the results of which were intensively discussed and revised.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"There are clearly eight orthographic words in the example given, but if one were doing syntactic analysis one would probably want to consider I'm to consist of two syntactic words, namely I and am.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"A different but supplementary perspective on discourse-based information structure is taken 11ventionalized patterns (e.g., order of informa by one of our partner projects, which is inter tion in news reports).","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"of Tokens MENERGI 318 160,000 200 180,000 IdentiFinder \u00e2\u0080\u0093 650,000 \u00e2\u0080\u0093 790,000 MENE \u00e2\u0080\u0093 \u00e2\u0080\u0093 350 321,000 Table 4: Training Data MUC7 test accuracy.2 For MUC6, the reduction in error due to global features is 27%, and for MUC7,14%.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"The segmentation chosen is the best path through the WFST, shown in (d).","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"And indeed, converging on annotation guidelines is even more difficult than it is with co-reference.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"For each domain, phrases which contain the same keyword are gathered to build a set of phrases (Step 3).","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"Model Overview The model starts by generating a tag assignment T for each word type in a vocabulary, assuming one tag per word.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Note that Chang, Chen, and Chen (1991), in addition to word-frequency information, include a constraint-satisfication model, so their method is really a hybrid approach.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"To make the projection practical, we rely on the twelve universal part-of-speech tags of Petrov et al. (2011).","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Many hanzi have more than one pronunciation, where the correct.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"It is possible one could produce better models by introducing features describing constituents and their contexts because one parser could be much better than the majority of the others in particular situations.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"If one system is perfect, another has slight flaws and the third more flaws, a judge is inclined to hand out judgements of 5, 4, and 3.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"Also, the method of using keywords rules out phrases which don\u00e2\u0080\u0099t contain popular words in the domain.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"For each feature type f and tag t, a multinomial \u00cf\u0088tf is drawn from a symmetric Dirichlet distribution with concentration parameter \u00ce\u00b2.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"This means that the rules in our grammar are of two kinds: (a) syntactic rules relating nonterminals to a sequence of non-terminals and\/or PoS tags, and (b) lexical rules relating PoS tags to lattice arcs (lexemes).","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"In order to solve this problem, a parse tree is needed to understand that \u00e2\u0080\u009cLotus\u00e2\u0080\u009d is not the object of \u00e2\u0080\u009cestimates\u00e2\u0080\u009d.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"We developed a first version of annotation guidelines for co-reference in PCC (Gross 2003), which served as basis for annotating the core corpus but have not been empirically evaluated for inter-annotator agreement yet.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"Skipped (K): The translation of up to one word may be postponed . Verb (V): The translation of up to two words may be anticipated.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"The system will then look for sequences of initial capitalized words that match the acronyms found in the whole document.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"For instance, for TTS it is necessary to know that a particular sequence of hanzi is of a particular category because that knowl\u00c2\u00ad edge could affect the pronunciation; consider, for example the issues surrounding the pronunciation of ganl I qian2 discussed in Section 1.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"In both cases, SRILM walks its trie an additional time to minimize context as mentioned in Section 4.1.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Sometimes, however, these beliefs can be contradictory.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"We use N(u) to denote the neighborhood of vertex u, and fixed n = 5 in our experiments.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"For instance, in the recent IWSLT evaluation, first fluency annotations were solicited (while withholding the source sentence), and then adequacy annotations.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"Put another way, the minimum of Equ.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"handled given appropriate models.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"Our results suggest that it is possible to learn accurate POS taggers for languages which do not have any annotated data, but have translations into a resource-rich language.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"We paid particular attention to minimize the number of free parameters, and used the same hyperparameters for all language pairs, rather than attempting language-specific tuning.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Values in the trie are minimally sized at the bit level, improving memory consumption over trie implementations in SRILM, IRSTLM, and BerkeleyLM.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Each visited entry wni stores backoff b(wni ).","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"21 In Chinese, numerals and demonstratives cannot modify nouns directly, and must be accompanied by.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"(e.g., N.Y. would contribute this feature, IBM would not). nonalpha=x Appears if the spelling contains any characters other than upper or lower case letters.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Obviously, the presence of a title after a potential name N increases the probability that N is in fact a name.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"The record for wn1 stores the offset at which its extensions begin.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Clearly it is possible to write a rule that states that if an analysis Modal+ Verb is available, then that is to be preferred over Noun+ Verb: such a rule could be stated in terms of (finite-state) local grammars in the sense of Mohri (1993).","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"logical rules, and personal names; the transitive closure of the resulting machine is then computed.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"This research was supported in part by the Defense Advanced Research Projects Agency as part of the Translingual Information Detection, Extraction and Summarization (TIDES) program, under Grant N66001001-18917 from the Space and Naval Warfare Systems Center, San Diego, and by the National Science Foundation under Grant IIS00325657.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"8 Volume 1, English language, 19961997, Format version 1, correction level 0 An ap ho r T e r r o r i s m R e c Pr F D i s a s t e r s R e c Pr F De f. NP s Pro no uns .43 .79 .55 .50 .72 .59 .42 .91 .58 .42 .82 .56 Tot al .46 .76 .57 .42 .87 .57 Table 2: General Knowledge Sources Table 4: Individual Performance of KSs for Terrorism Table 3: General + Contextual Role Knowledge Sources larger MUC4 and Reuters corpora.9 4.2 Experiments.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"For example, both a chair and a truck would be labeled as artifacts, but this does not at all suggest that they are coreferent.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"Because of their size, the examples (Figures 2 to 4) appear at the end of the paper.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"hanzi in the various name positions, derived from a million names.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Lazy mapping reduces memory requirements by loading pages from disk only as necessary.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"We define a symmetric similarity function K(uZ7 uj) over two foreign language vertices uZ7 uj E Vf based on the co-occurrence statistics of the nine feature concepts given in Table 1.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"In addition to the Europarl test set, we also collected 29 editorials from the Project Syndicate website2, which are published in all the four languages of the shared task.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Suffixes and Prefixes: This group contains only two features: Corporate-Suffix and Person-Prefix.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"In the first stage, we run a single step of label propagation, which transfers the label distributions from the English vertices to the connected foreign language vertices (say, Vf\ufffd) at the periphery of the graph.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Figure 5 shows how this model is implemented as part of the dictionary WFST.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"A Person-Prefix-List is compiled in an analogous way.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Linguistic intuitions like those in the previous section inform language-specific annotation choices.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"The equation for sampling a single type-level assignment Ti is given by, 0.2 0 5 10 15 20 25 30 Iteration Figure 2: Graph of the one-to-one accuracy of our full model (+FEATS) under the best hyperparameter setting by iteration (see Section 5).","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"Table 9 shows that MADA produces a high quality segmentation, and that the effect of cascading segmentation errors on parsing is only 1.92% F1.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"In German, the verbgroup usually consists of a left and a right verbal brace, whereas in English the words of the verbgroup usually form a sequence of consecutive words.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"If the token is the first word of a sentence, then this feature is set to 1.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Chris Dyer integrated the code into cdec.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Space- or punctuation-delimited * 700 Mountain Avenue, 2d451, Murray Hill, NJ 07974, USA.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"The bias of automatic methods in favor of statistical systems seems to be less pronounced on out-of-domain test data.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"We are given a source string fJ 1 = f1:::fj :::fJ of length J, which is to be translated into a target string eI 1 = e1:::ei:::eI of length I. Among all possible target strings, we will choose the string with the highest probability: ^eI 1 = arg max eI 1 fPr(eI 1jfJ 1 )g = arg max eI 1 fPr(eI 1) Pr(fJ 1 jeI 1)g : (1) The argmax operation denotes the search problem, i.e. the generation of the output sentence in the target language.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"1.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"The translation of one position in the source sentence may be postponed for up to L = 3 source positions, and the translation of up to two source positions may be anticipated for at most R = 10 source positions.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"86 78.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"Hence, the resulting trees will no longer have two branches of equal size.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"The gender, number, and scoping KSs eliminate candidates from consideration.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"Let say, if we find one system doing better on 20 of the blocks, and worse on 80 of the blocks, is it significantly worse?","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"The pseudo-code describing the algorithm is given in Fig.","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"The choice of the genre commentary resulted from the fact that an investigation of rhetorical structure, its interaction with other aspects of discourse structure, and the prospects for its automatic derivation are the key motivations for building up the corpus.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"Hence, we use the bootstrap resampling method described by Koehn (2004).","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"Fig.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"ogy (Koskenniemi 1983; Antworth 1990; Tzoukermann and Liberman 1990; Karttunen, Kaplan, and Zaenen 1992; Sproat 1992); we represent the fact that ir, attaches to nouns by allowing t:-transitions from the final states of all noun entries, to the initial state of the sub-WFST representing f,.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"Morphological Analyzer Ideally, we would use an of-the-shelf morphological analyzer for mapping each input token to its possible analyses.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":", December, then the feature MonthName is set to 1.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"Participants were also provided with two sets of 2,000 sentences of parallel text to be used for system development and tuning.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"For Arabic we M o d e l S y s t e m L e n g t h L e a f A n c e s t o r Co rpu s Sent Exact E v a l b L P LR F1 T a g % B a s e l i n e 7 0 St an for d (v 1.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"We will refer to the semantic classes that co-occur with a caseframe as the semantic expectations of the caseframe.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"Annotators argued for the importance of having correct and even multiple references.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"For each pair of judges consider one judge as the standard,.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"Systems that generally do worse than others will receive a negative one.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"There is a sizable literature on Chinese word segmentation: recent reviews include Wang, Su, and Mo (1990) and Wu and Tseng (1993).","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"In this work, we take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"For each pair we also record the context, i.e. the phrase between the two NEs (Step1).","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"Equation 2 is an estimate of the conditional probability of the label given the feature, P(yjx).","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"Note that since all English vertices were extracted from the parallel text, we will have an initial label distribution for all vertices in Ve.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"This process produces a large set of caseframes coupled with a list of the noun phrases that they extracted.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"For each 13 Of course, this weighting makes the PCFG an improper distribution.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"12 One class of full personal names that this characterization does not cover are married women's names.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"Let say, if we find one system doing better on 20 of the blocks, and worse on 80 of the blocks, is it significantly worse?","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"The sentence-selection approach is crude in that it imposes a binary distinction between useful and non-useful parts of OUT.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"3 Techniques for automatic vocalization have been studied (Zitouni et al., 2006; Habash and Rambow, 2007).","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"We computed BLEU scores for each submission with a single reference translation.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"When dropping the top and bottom 2.5% the remaining BLEU scores define the range of the confidence interval.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"The attachment in such cases encompasses a long distance dependency that cannot be captured by Markovian processes that are typically used for morphological disambiguation.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"The discovered paraphrases can be a big help to reduce human labor and create a more comprehensive pattern set.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"1.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"(1998) did make use of information from the whole document.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Then each arc of D maps either from an element of H to an element of p, or from E-i.e., the empty string-to an element of P. More specifically, each word is represented in the dictionary as a sequence of arcs, starting from the initial state of D and labeled with an element 5 of Hxp, which is terminated with a weighted arc labeled with an element of Ex P. The weight represents the estimated cost (negative log probability) of the word.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"In our model there are no such hyper-parameters, and the performance is the result of truly joint disambiguation. sults.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"For a trigram language model, the partial hypotheses are of the form (e0; e; C; j).","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","label":1},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"The terms on the right-hand-side denote the type-level and token-level probability terms respectively.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"Finally, the assignments of PoS tags to OOV segments is subject to language specific constraints relative to the token it was originated from.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"The effect of a second reference resolution classifier is not entirely the same as that of global features.","label":0},{"source_text":"There is no global pruning.","target_text":"For the experiments, we use a simple preprocessing step.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"The CoBoost algorithm just described is for the case where there are two labels: for the named entity task there are three labels, and in general it will be useful to generalize the CoBoost algorithm to the multiclass case.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"We dropped, however, one of the languages, Finnish, partly to keep the number of tracks manageable, partly because we assumed that it would be hard to find enough Finnish speakers for the manual evaluation.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"We evaluate our model on seven languages exhibiting substantial syntactic variation.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"(1) CEO of McCann . . .","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"As a result, Arabic sentences are usually long relative to English, especially after Length English (WSJ) Arabic (ATB) \u00e2\u0089\u00a4 20 41.9% 33.7% \u00e2\u0089\u00a4 40 92.4% 73.2% \u00e2\u0089\u00a4 63 99.7% 92.6% \u00e2\u0089\u00a4 70 99.9% 94.9% Table 2: Frequency distribution for sentence lengths in the WSJ (sections 2\u00e2\u0080\u009323) and the ATB (p1\u00e2\u0080\u00933).","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"In this paper we study the problem of using a parallel corpus from a background domain (OUT) to improve performance on a target domain (IN) for which a smaller amount of parallel training material\u2014though adequate for reasonable performance\u2014is also available.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"In previous work (Bean and Riloff, 1999), we developed an unsupervised learning algorithm that automatically recognizes definite NPs that are existential without syntactic modification because their meaning is universally understood.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"3.2 Results.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"Consequently, we cut their evidence values in half to lessen their influence.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"For example, suppose one is building a ITS system for Mandarin Chinese.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"The BLEU score has been shown to correlate well with human judgement, when statistical machine translation systems are compared (Doddington, 2002; Przybocki, 2004; Li, 2005).","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"3.2 Inter-annotator Agreement.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"It may be more realistic to replace the second criteria with a softer one, for example (Blum and Mitchell 98) suggest the alternative Alternatively, if Ii and 12 are probabilistic learners, it might make sense to encode the second constraint as one of minimizing some measure of the distance between the distributions given by the two learners.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"Table 2 shows results for both settings and all methods described in sections 2 and 3.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"encodes the one tag per word constraint and is uni form over type-level tag assignments.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"The anaphor and antecedent appear in boldface.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"The availability of comparable corpora is limited, which is a significant limitation on the approach.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"This section describes an algorithm based on boosting algorithms, which were previously developed for supervised machine learning problems.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"In our situation, the competing hypotheses are the possible antecedents for an anaphor.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"Top 5 Bot to m 5 Go ld NN P NN JJ CD NN S RB S PD T # \u00e2\u0080\u009d , 1T W CD W RB NN S VB N NN PR P$ W DT : MD . +P RI OR CD JJ NN S WP $ NN RR B- , $ \u00e2\u0080\u009d . +F EA TS JJ NN S CD NN P UH , PR P$ # . \u00e2\u0080\u009c Table 5: Type-level English POS Tag Ranking: We list the top 5 and bottom 5 POS tags in the lexicon and the predictions of our models under the best hyperparameter setting.","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"The kind of annotation work presented here would clearly benefit from the emergence of standard formats and tag sets, which could lead to sharable resources of larger size.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"To initialize the graph we tag the English side of the parallel text using a supervised model.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"Part-of-speech (POS) tag distributions are known to exhibit sparsity \u00e2\u0080\u0094 a word is likely to take a single predominant tag in a corpus.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"The parser switching oracle is the upper bound on the accuracy that can be achieved on this set in the parser switching framework.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"5.2 Setup.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"orthographic words are thus only a starting point for further analysis and can only be regarded as a useful hint at the desired division of the sentence into words.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"We of course also fail to identify, by the methods just described, given names used without their associated family name.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"The entries in this table can be compared with those of Table 3 to see how the performance of the combining techniques degrades in the presence of an inferior parser.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"If two phrases can be used to express the same relationship within an information extraction application (\u00e2\u0080\u009cscenario\u00e2\u0080\u009d), these two phrases are paraphrases.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"As observed by Kahane et al. (1998), any (nonprojective) dependency graph can be transformed into a projective one by a lifting operation, which replaces each non-projective arc wj wk by a projective arc wi \u2014* wk such that wi \u2014*\u2217 wj holds in the original graph.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"6.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"The system of Berg-Kirkpatrick et al.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Our system fails in (a) because of$ shenl, a rare family name; the system identifies it as a family name, whereas it should be analyzed as part of the given name.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"For example, BABAR learned that agents that \u00e2\u0080\u009cassassinate\u00e2\u0080\u009d or \u00e2\u0080\u009cinvestigate a cause\u00e2\u0080\u009d are usually humans or groups (i.e., organizations).","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"This extends previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and using a simpler training procedure.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"We model po(s|t) using a MAP criterion over weighted phrase-pair counts: and from the similarity to (5), assuming y = 0, we see that w\u03bb(s, t) can be interpreted as approximating pf(s, t)\/po(s, t).","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"To this end, we picked 100 sentences at random containing 4,372 total hanzi from a test corpus.14 (There were 487 marks of punctuation in the test sentences, including the sentence-final periods, meaning that the average inter-punctuation distance was about 9 hanzi.)","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"Smith estimates Lotus will make a profit this quarter\u00e2\u0080\u00a6\u00e2\u0080\u009d, our system extracts \u00e2\u0080\u009cSmith esti mates Lotus\u00e2\u0080\u009d as an instance.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"For this experiment, we compare our model with the uniform tag assignment prior (1TW) with the learned prior (+PRIOR).","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"First of all, most previous articles report perfor\u00c2\u00ad mance in terms of a single percent-correct score, or else in terms of the paired measures of precision and recall.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"(Levinger et al., 1995; Goldberg et al., ; Adler et al., 2008)) will make the parser more robust and suitable for use in more realistic scenarios.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"So, 1: f, xue2shengl+men0 (student+PL) 'students' occurs and we estimate its cost at 11.43; similarly we estimate the cost of f, jiang4+men0 (general+PL) 'generals' (as in 'J' f, xiao3jiang4+men0 'little generals'), at 15.02.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"3.1 General Knowledge Sources.","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"However, while Dienes and Dubey recognize empty categories in a pre-processing step and only let the parser find their antecedents, we use the parser both to detect dislocated dependents and to predict either the type or the location of their syntactic head (or both) and use post-processing only to transform the graph in accordance with the parser\u2019s analysis.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"We present several variations for the lexical component P (T , W |\u00cf\u0088), each adding more complex pa rameterizations.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"In any event, to date, we have not compared different methods for deriving the set of initial frequency estimates.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"In the next section we present an alternative approach that builds two classifiers while attempting to satisfy the above constraints as much as possible.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"When aligning the words in parallel texts (for language pairs like SpanishEnglish, French-English, ItalianGerman,...), we typically observe a strong localization effect.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"M must enter a universal state and check that each of the k constant substrings are in the appropriate place (as determined by the contents of the first 2k work tapes) on the input tape.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"We applied the AutoSlog system (Riloff, 1996) to our unannotated training texts to generate a set of extraction patterns for each domain.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"4.4 Chinese Personal Names.","label":0},{"source_text":"Here we present two algorithms.","target_text":"For example, ... fraud related to work on a federally funded sewage plant in Georgia In this case, Georgia is extracted: the NP containing it is a complement to the preposition in; the PP headed by in modifies the NP a federally funded sewage plant, whose head is the singular noun plant.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"To optimize left-to-right queries, we extend state to store backoff information: where m is the minimal context from Section 4.1 and b is the backoff penalty.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"30 16.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"2.2.4 Semantic Caseframe Expectations The third type of contextual role knowledge learned by BABAR is Semantic Caseframe Expectations.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"(2006).","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"Each parse is converted into a set of constituents represented as a tuples: (label, start, end).","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"It uses a maximum entropy framework and classifies each word given its features.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"For MUC7, there are also no published results on systems trained on only the official training data of 200 aviation disaster articles.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"For the English RST-annotated corpus that is made available via LDC, his corresponding result is 62%.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"This has the effect of randomly permuting vocabulary identifiers, meeting the requirements of interpolation search when vocabulary identifiers are used as keys.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"But gold segmentation is not available in application settings, so a segmenter and parser are arranged in a pipeline.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"There is twice as much language modelling data, since training data for the machine translation system is filtered against sentences of length larger than 40 words.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"We will describe the evaluation of such clusters in the next subsection.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"The precision and recall measures (described in more detail in Section 3) used in evaluating Treebank parsing treat each constituent as a separate entity, a minimal unit of correctness.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"However, we note that the final conditional estimates p(s|t) from a given phrase table maximize the likelihood of joint empirical phrase pair counts over a word-aligned corpus.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"We consider the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"For comparison to information-retrieval inspired baselines, eg (L\u00a8u et al., 2007), we select sentences from OUT using language model perplexities from IN.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Still, for both human and automatic rhetorical analysis, connectives are the most important source of surface information.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"Different sentence structure and rich target language morphology are two reasons for this.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"The algorithm can be viewed as heuristically optimizing an objective function suggested by (Blum and Mitchell 98); empirically it is shown to be quite successful in optimizing this criterion.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Automatic scores are computed on a larger tested than manual scores (3064 sentences vs. 300\u2013400 sentences). collected manual judgements, we do not necessarily have the same sentence judged for both systems (judges evaluate 5 systems out of the 8\u201310 participating systems).","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"5.2 Discussion.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"When finished, the whole material is written into an XML-structured annotation file.","label":0},{"source_text":"Here we present two algorithms.","target_text":"Put another way, the minimum of Equ.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"However, there is a crucial difference: the morphological probabilities in their model come from discriminative models based on linear context.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"However, there will remain a large number of words that are not readily adduced to any produc\u00c2\u00ad tive pattern and that would simply have to be added to the dictionary.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"The larger sets are more accurate than the small sets.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"A named entity recognizer (NER) is useful in many NLP applications such as information extraction, question answering, etc. On its own, a NER can also provide users who are looking for person or organization names with quick information.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"2.2.4 Semantic Caseframe Expectations The third type of contextual role knowledge learned by BABAR is Semantic Caseframe Expectations.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"(c) After they blindfolded the men...","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"For graph propagation, the hyperparameter v was set to 2 x 10\u22126 and was not tuned.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"4.1 The Task and the Corpus.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"\u00e2\u0088\u0097NP NP PP R) and \u00e2\u0088\u0097NP NP ADJP R) are both iDafa attachment.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"To lower the barrier of entrance to the competition, we provided a complete baseline MT system, along with data resources.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"The average individual parser accuracy was reduced by more than 5% when we added this new parser, but the precision of the constituent voting technique was the only result that decreased significantly.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"87 Table 7: Test set results.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"While the semantic aspect of radicals is by no means completely predictive, the semantic homogeneity of many classes is quite striking: for example 254 out of the 263 examples (97%) of the INSECT class listed by Wieger (1965, 77376) denote crawling or invertebrate animals; similarly 21 out of the 22 examples (95%) of the GHOST class (page 808) denote ghosts or spirits.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"The Potsdam Commentary Corpus","label":1},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"For each domain, we created a semantic dictionary by doing two things.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"First, a non-anaphoric NP classifier identifies definite noun phrases that are existential, using both syntactic rules and our learned existential NP recognizer (Bean and Riloff, 1999), and removes them from the resolution process.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"4 65.9 48.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"Since the inclusion of out-ofdomain test data was a very late decision, the participants were not informed of this.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Overall, language modeling significantly impacts decoder performance.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"Morphological Analyzer Ideally, we would use an of-the-shelf morphological analyzer for mapping each input token to its possible analyses.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"(Blum and Mitchell 98) describe learning in the following situation: X = X1 X X2 where X1 and X2 correspond to two different &quot;views&quot; of an example.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"We also performed experiments to evaluate the impact of each type of contextual role knowledge separately.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"The standard measures for evaluating Penn Treebank parsing performance are precision and recall of the predicted constituents.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"Our full model (\u201cWith LP\u201d) outperforms the unsupervised baselines and the \u201cNo LP\u201d setting for all languages.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"While we had up to 11 submissions for a translation direction, we did decide against presenting all 11 system outputs to the human judge.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"The normalization on a per-judge basis gave very similar ranking, only slightly less consistent with the ranking from the pairwise comparisons.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"Unsupervised Part-of-Speech Tagging with Bilingual Graph-Based Projections","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"For every sequence of initial capitalized words, its longest substring that occurs in the same document as a sequence of initCaps is identified.","label":0},{"source_text":"There is no global pruning.","target_text":"Here, we process only full-form words within the translation procedure.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"The last affix in the list is the nominal plural f, men0.20 In the table are the (typical) classes of words to which the affix attaches, the number found in the test corpus by the method, the number correct (with a precision measure), and the number missed (with a recall measure).","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"The label for a test example with features x is then defined as In this paper we define h(x, y) as the following function of counts seen in training data: Count(x,y) is the number of times feature x is seen with label y in training data, Count(x) = EyEy Count(x, y). a is a smoothing parameter, and k is the number of possible labels.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"irL as the product of the probability estimate for i\u00c2\u00a5JJ1l., and the probability estimate just derived for unseen plurals in ir,: p(i\u00c2\u00a51J1l.ir,) p(i\u00c2\u00a51J1l.)p(unseen(f,)).","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"We attain these results using several optimizations: hashing, custom lookup tables, bit-level packing, and state for left-to-right query patterns.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"In Chinese text, individual characters of the script, to which we shall refer by their traditional name of hanzi,Z are written one after another with no intervening spaces; a Chinese sentence is shown in Figure 1.3 Partly as a result of this, the notion \"word\" has never played a role in Chinese philological tradition, and the idea that Chinese lacks any\u00c2\u00ad thing analogous to words in European languages has been prevalent among Western sinologists; see DeFrancis (1984).","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"For example, if is found in the list of person first names, the feature PersonFirstName is set to 1.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"na me =>2 ha nzi fa mi ly 1 ha nzi gi ve n 6.1 ha nzi fa mi ly => ha nz ii 7.2 ha nzi fa mi ly => ha nzi i ha nz ij 8.1 ha nzi gi ve n => ha nz ii 9.2 ha nzi giv en => ha nzi i ha nz ij The difficulty is that given names can consist, in principle, of any hanzi or pair of hanzi, so the possible given names are limited only by the total number of hanzi, though some hanzi are certainly far more likely than others.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"In most cases, however, these expansions come with a steep increase in model complexity, with respect to training procedure and inference time.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"We apply a beam search concept as in speech recognition.","label":1},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"971,746 sentences of New York Times text were parsed using the parser of (Collins 96).1 Word sequences that met the following criteria were then extracted as named entity examples: whose head is a singular noun (tagged NN).","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"However, such an approach requires a number of hand-coded systems, which may not be available in languages other than English.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"E.g. when 'Zahnarzttermin' is aligned to dentist's, the extended lexicon model might learn that 'Zahnarzttermin' actuallyhas to be aligned to both dentist's and ap pointment.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"In addition to a heuristic based on decision list learning, we also presented a boosting-like framework that builds on ideas from (Blum and Mitchell 98).","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Sentence (2) and (3) help to disambiguate one way or the other.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"We believe it is natural for authors to use abbreviations in subsequent mentions of a named entity (i.e., first \u00e2\u0080\u009cPresident George Bush\u00e2\u0080\u009d then \u00e2\u0080\u009cBush\u00e2\u0080\u009d).","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"The last four columns in Table 3 show the distribution of nonprojective arcs with respect to the number of lifts required.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"( b ) s u p p o r t s c a n d i d a t e i f s e l e c t e d s e m a n t i c t a g s m a t c h t h o s e o f t h e a n a p h o r . Le xic al computes degree of lexical overlap b e t w e e n t h e c a n d i d a t e a n d t h e a n a p h o r . Re cen cy computes the relative distance between the c a n d i d a t e a n d t h e a n a p h o r . Sy nR ole computes relative frequency with which the c a n d i d a t e \u00e2\u0080\u0099 s s y n t a c t i c r o l e o c c u r s i n r e s o l u t i o n s . Figure 4: General Knowledge Sources The Lexical KS returns 1 if the candidate and anaphor are identical, 0.5 if their head nouns match, and 0 otherwise.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Our code is thread-safe, and integrated into the Moses, cdec, and Joshua translation systems.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"Each token may admit multiple analyses, each of which a sequence of one or more lexemes (we use li to denote a lexeme) belonging a presupposed Hebrew lexicon LEX.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Like in the co-reference annotation, G\u00c2\u00a8otze\u00e2\u0080\u0099s proposal has been applied by two annotators to the core corpus but it has not been systematically evaluated yet.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Moreover, they are used as substantives much 2 Unlike machine translation, constituency parsing is not significantly affected by variable word order.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"Like verbs, maSdar takes arguments and assigns case to its objects, whereas it also demonstrates nominal characteristics by, e.g., taking determiners and heading iDafa (Fassi Fehri, 1993).","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Microsoft\u2019s approach uses dependency trees, others use hierarchical phrase models.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"We show translation results for three approaches: the monotone search (MonS), where no word reordering is allowed (Tillmann, 1997), the quasimonotone search (QmS) as presented in this paper and the IBM style (IbmS) search as described in Section 3.2.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"This is manifest in the lexical choices but 1 www.coli.unisb.de\/\u00e2\u0088\u00bcthorsten\/tnt\/ Dagmar Ziegler is up to her neck in debt.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"We observe similar trends when using another measure \u00e2\u0080\u0093 type-level accuracy (defined as the fraction of words correctly assigned their majority tag), according to which La ng ua ge M etr ic B K 10 E M B K 10 L B F G S G 10 F EA T S B es t F EA T S M ed ia n E ng lis h 1 1 m 1 4 8 . 3 6 8 . 1 5 6 . 0 7 5 . 5 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 0 . 9 6 6 . 4 4 7 . 8 6 6 . 4 D an is h 1 1 m 1 4 2 . 3 6 6 . 7 4 2 . 6 5 8 . 0 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 2 . 1 6 1 . 2 4 3 . 2 6 0 . 7 D ut ch 1 1 m 1 5 3 . 7 6 7 . 0 5 5 . 1 6 4 . 7 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 6 . 4 6 9 . 0 5 1 . 5 6 7 . 3 Po rtu gu es e 1 1 m 1 5 0 . 8 7 5 . 3 4 3 . 2 7 4 . 8 44 .5 69 .2 6 4 . 1 7 4 . 5 5 6 . 5 7 0 . 1 S pa ni sh 1 1 m 1 \u00e2\u0080\u0093 \u00e2\u0080\u0093 4 0 . 6 7 3 . 2 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 8 . 3 6 8 . 9 5 0 . 0 5 7 . 2 Table 4: Comparison of our method (FEATS) to state-of-the-art methods.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"It is chosen such that the decisions it made in including or excluding constituents are most probable under the models for all of the parsers.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"The problem of &quot;noise&quot; items that do not fall into any of the three categories also needs to be addressed.","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"The dev corpus was taken from the NIST05 evaluation set, augmented with some randomly-selected material reserved from the training set.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"As noted for the perplexity task, we do not expect cache to grow substantially with model size, so RandLM remains a low-memory option.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"We realize the importance of paraphrase; however, the major obstacle is the construction of paraphrase knowledge.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"!!:\\ :yu2 e:_nc [::!!:zen3 l!f :moO t:_adv il!:shuot ,:_vb i i i 1 \u00e2\u0080\u00a2 10.03 13...","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"To resolve the anaphor, we survey the final belief values assigned to each candidate\u00e2\u0080\u0099s singleton set.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"In focus is in particular the correlation with rhetorical structure, i.e., the question whether specific rhetorical relations \u00e2\u0080\u0094 or groups of relations in particular configurations \u00e2\u0080\u0094 are signalled by speakers with prosodic means.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"Given the bilingual graph described in the previous section, we can use label propagation to project the English POS labels to the foreign language.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"Figure 1 provides some statistics about this corpus.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Annotators suggested that long sentences are almost impossible to judge.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Timing is based on plentiful memory.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Because the Bikel parser has been parameter- ized for Arabic by the LDC, we do not change the default model settings.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Table 2 shows single-threaded results, mostly for comparison to IRSTLM, and Table 3 shows multi-threaded results.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"In our work, we demonstrate that using a simple na\u00c2\u00a8\u00c4\u00b1veBayes approach also yields substantial performance gains, without the associated training complexity.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"The second row is the accuracy of the best of the three parsers.'","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"Thus, provided at least this amount of IN data is available\u2014as it is in our setting\u2014adapting these weights is straightforward.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"If no candidate satisfies this condition (which is often the case), then the anaphor is left unresolved.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"Specifically, for the ith word type, the set of token-level tags associated with token occurrences of this word, denoted t(i), must all take the value Ti to have nonzero mass. Thus in the context of Gibbs sampling, if we want to block sample Ti with t(i), we only need sample values for Ti and consider this setting of t(i).","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"This is similar to stacking the different feature instantiations into long (sparse) vectors and computing the cosine similarity between them.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"As is standard, we report the greedy one-to-one (Haghighi and Klein, 2006) and the many-to-one token-level accuracy obtained from mapping model states to gold POS tags.","label":0},{"source_text":"Here we present two algorithms.","target_text":"This paper discusses the use of unlabeled examples for the problem of named entity classification.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Se ma nti c (a) filters candidate if its semantic tags d o n \u00e2\u0080\u0099 t i n t e r s e c t w i t h t h o s e o f t h e a n a p h o r .","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"The corresponding token words w are drawn conditioned on t and \u00ce\u00b8.2 Our full generative model is given by: K P (\u00cf\u0086, \u00ce\u00b8|T , \u00ce\u00b1, \u00ce\u00b2) = n (P (\u00cf\u0086t|\u00ce\u00b1)P (\u00ce\u00b8t|T , \u00ce\u00b1)) t=1 The transition distribution \u00cf\u0086t for each tag t is drawn according to DIRICHLET(\u00ce\u00b1, K ), where \u00ce\u00b1 is the shared transition and emission distribution hyperparameter.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"This is most severe with RandLM in the multi-threaded case, where each thread keeps a separate cache, exceeding the original model size.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"The semantic agreement KS eliminates some candidates, but also provides positive evidence in one case: if the candidate and anaphor both have semantic tags human, company, date, or location that were assigned via NER or the manually labeled dictionary entries.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"However, if we consider precision, recall and Fmeasure on non-projective dependencies only, as shown in Table 6, some differences begin to emerge.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"The approach gains leverage from natural redundancy in the data: for many named-entity instances both the spelling of the name and the context in which it appears are sufficient to determine its type.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"To be short, we omit the target words e; e0 in the formulation of the search hypotheses.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"Also, the argument has been made that machine translation performance should be evaluated via task-based evaluation metrics, i.e. how much it assists performing a useful task, such as supporting human translators or aiding the analysis of texts.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"For example, the phrase \u00e2\u0080\u009c's New York-based trust unit,\u00e2\u0080\u009d is not a paraphrase of the other phrases in the \u00e2\u0080\u009cunit\u00e2\u0080\u009d set.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Furthermore, by inverting the transducer so that it maps from phonemic transcriptions to hanzi sequences, one can apply the segmenter to other problems, such as speech recognition (Pereira, Riley, and Sproat 1994).","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"3 The Coreference Resolution Model.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"\u00e2\u0080\u009d, because Bush is the first word, the initial caps might be due to its position (as in \u00e2\u0080\u009cThey put a freeze on . . .","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Others depend upon various lexical heuris\u00c2\u00ad tics: for example Chen and Liu (1992) attempt to balance the length of words in a three-word window, favoring segmentations that give approximately equal length for each word.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"genitiveMark indicates recursive NPs with a indefinite nominal left daughter and an NP right daughter.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"MUC7 has also seen hybrids of statistical NERs and hand-coded systems (Mikheev et al., 1998; Borthwick, 1999), notably Mikheev' s system, which achieved the best performance of 93.39% on the official NE test data.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"2.6 Co-reference.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"encodes the one tag per word constraint and is uni form over type-level tag assignments.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"Frequency is calculated by counting the number of distinct previous tokens that each token has (e.g., if Electric Corp. is seen 3 times, and Manufacturing Corp. is seen 5 times during training, and Corp. is not seen with any other preceding tokens, then the \u00e2\u0080\u009cfrequency\u00e2\u0080\u009d of Corp. is 2).","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"As expected, the vanilla HMM trained with EM performs the worst.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"The techniques we develop can be extended in a relatively straightforward manner to the more general case when OUT consists of multiple sub-domains.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"The Berkeley parser gives state-of-the-art performance for all metrics.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"We will also directly compare with a baseline similar to the Matsoukas et al approach in order to measure the benefit from weighting phrase pairs (or ngrams) rather than full sentences.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"((fii, Q2, Pa) , \u2014\u25a0 (01, i32, 03) , e),(02,e) Oa, en) The path complexity of the tee set generated by a MCTAG is not necessarily context-free.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"16 As one reviewer points out, one problem with the unigram model chosen here is that there is still a. tendency to pick a segmentation containing fewer words.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"(2009).","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models (e.g., (Grosz et al., 1995; Grosz and Sidner, 1998)), syntactic algorithms (e.g., (Hobbs, 1978; Lappin and Le- ass, 1994)), and supervised machine learning systems (Aone and Bennett, 1995; McCarthy and Lehnert, 1995; Ng and Cardie, 2002; Soon et al., 2001).","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"18 We are grateful to ChaoHuang Chang for providing us with this set.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"The problem is a binary classification problem.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"Formally, the DempsterShafer theory defines a probability density function m(S), where S is a set of hypotheses.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"Each lattice arc corresponds to a segment and its corresponding PoS tag, and a path through the lattice corresponds to a specific morphological segmentation of the utterance.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Thus, we feel fairly confident that for the examples we have considered from Gan's study a solution can be incorporated, or at least approximated, within a finite-state framework.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"(2009) study related but different multilingual grammar and tagger induction tasks, where it is assumed that no labeled data at all is available.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"The maximum entropy framework estimates probabilities based on the principle of making as few assumptions as possible, other than the constraints imposed.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Dynamic programming efficiently scores many hypotheses by exploiting the fact that an N-gram language model conditions on at most N \u2212 1 preceding words.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Among the IS-units, the referring expressions are marked as such and will in the second phase receive a label for cognitive status (active, accessible- text, accessible-situation, inferrable, inactive).","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Though we do not directly compare state implementations, performance metrics in Table 1 indicate our overall method is faster.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"The wounds are still healing.), entity-attribute (e.g., She 2001), who determined that in their corpus of German computer tests, 38% of relations were lexically signalled.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"When aligning the words in parallel texts (for language pairs like SpanishEnglish, French-English, ItalianGerman,...), we typically observe a strong localization effect.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"na me =>2 ha nzi fa mi ly 1 ha nzi gi ve n 6.1 ha nzi fa mi ly => ha nz ii 7.2 ha nzi fa mi ly => ha nzi i ha nz ij 8.1 ha nzi gi ve n => ha nz ii 9.2 ha nzi giv en => ha nzi i ha nz ij The difficulty is that given names can consist, in principle, of any hanzi or pair of hanzi, so the possible given names are limited only by the total number of hanzi, though some hanzi are certainly far more likely than others.","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"For each token, three types of features may be taken into account: the word form; the part-of-speech assigned by an automatic tagger; and labels on previously assigned dependency arcs involving the token \u2013 the arc from its head and the arcs to its leftmost and rightmost dependent, respectively.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"(2008) reported agreement between the teams (measured with Evalb) at 93.8% F1, the level of the CTB.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Since Dane\u00cb\u0087s\u00e2\u0080\u0099 proposals of \u00e2\u0080\u0098thematic development patterns\u00e2\u0080\u0099, a few suggestions have been made as to the existence of a level of discourse structure that would predict the information structure of sentences within texts.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"This has the potential drawback of increasing the number of features, which can make MERT less stable (Foster and Kuhn, 2009).","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"Our new approach, which is called quasi-monotone search, processes the source sentence monotonically, while explicitly taking into account the positions of the German verbgroup.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"For a trigram language model, the partial hypotheses are of the form (e0; e; C; j).","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"For each language, we took the same number of sentences from the bitext as there are in its treebank, and trained a supervised feature-HMM.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Both of these analyses are shown in Figure 4; fortunately, the correct analysis is also the one with the lowest cost, so it is this analysis that is chosen.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"The bias of automatic methods in favor of statistical systems seems to be less pronounced on out-of-domain test data.","label":1},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"4.2 A Sample Segmentation Using Only Dictionary Words Figure 4 shows two possible paths from the lattice of possible analyses of the input sentence B X:\u00c2\u00a5 .:.S:P:l 'How do you say octopus in Japanese?' previously shown in Figure 1.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"La ng ua ge # To ke ns # W or d Ty pe s # Ta gs E ng lis h D a ni s h D u tc h G e r m a n P or tu g u e s e S p a ni s h S w e di s h 1 1 7 3 7 6 6 9 4 3 8 6 2 0 3 5 6 8 6 9 9 6 0 5 2 0 6 6 7 8 8 9 3 3 4 1 9 1 4 6 7 4 9 2 0 6 1 8 3 5 6 2 8 3 9 3 7 2 3 2 5 2 8 9 3 1 1 6 4 5 8 2 0 0 5 7 4 5 2 5 1 2 5 4 2 2 4 7 4 1 Table 2: Statistics for various corpora utilized in experiments.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"For this example, since the sequence Even News Broadcasting Corp. only appears once in the document, its longest substring that occurs in the same document is News Broadcasting Corp. In this case, News has an additional feature of I begin set to 1, Broadcasting has an additional feature of I continue set to 1, and Corp. has an additional feature of I end set to 1.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"The transitive closure of the dictionary in (a) is composed with Id(input) (b) to form the WFST (c).","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"The rhetorical structure annotations of PCC have all been converted to URML.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"In addition to the optimizations specific to each datastructure described in Section 2, we implement several general optimizations for language modeling.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"Corporate Suffixes and Person Prefixes of Other Occurrences (CSPP): If McCann has been seen as Mr. McCann somewhere else in the document, then one would like to give person a higher probability than organization.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"m(S) represents the belief that the correct hypothesis is included in S. The model assumes that evidence also arrives as a probability density function (pdf) over sets of hypotheses.6 Integrating new evidence into the existing model is therefore simply a matter of defining a function to merge pdfs, one representing the current belief system and one representing the beliefs of the new evidence.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"The negative logarithm of t0 is reported.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"Domain adaptation is a common concern when optimizing empirical NLP applications.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Link phrases based on instance pairs Using NE instance pairs as a clue, we find links between sets of phrases.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"The tokens w are generated by token-level tags t from an HMM parameterized by the lexicon structure.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"The main disadvantage of manual evaluation is that it is time-consuming and thus too expensive to do frequently.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Like in the co-reference annotation, G\u00c2\u00a8otze\u00e2\u0080\u0099s proposal has been applied by two annotators to the core corpus but it has not been systematically evaluated yet.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"The use of ILP in learning the desired grammar significantly increases the computational complexity of this method.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"Specifically, we assume each word type W consists of feature-value pairs (f, v).","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"Across eight European languages, our approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"Equation 2 is an estimate of the conditional probability of the label given the feature, P(yjx).","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"Lexical rules are estimated in a similar manner.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"The test set included 2000 sentences from the Europarl corpus, but also 1064 sentences out-ofdomain test data.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Other kinds of productive word classes, such as company names, abbreviations (termed fijsuolxie3 in Mandarin), and place names can easily be 20 Note that 7 in E 7 is normally pronounced as leO, but as part of a resultative it is liao3..","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"The first row represents the average accuracy of the three parsers we combine.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"3.3 Evaluation Results.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"An examination of the subjects' bracketings confirmed that these instructions were satisfactory in yielding plausible word-sized units.","label":0},{"source_text":"Here we present two algorithms.","target_text":"A generative model was applied (similar to naive Bayes) with the three labels as hidden vanables on unlabeled examples, and observed variables on (seed) labeled examples.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"The approach uses both spelling and contextual rules.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"A minimal requirement for building a Chinese word segmenter is obviously a dictionary; furthermore, as has been argued persuasively by Fung and Wu (1994), one will perform much better at segmenting text by using a dictionary constructed with text of the same genre as the text to be segmented.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"We thank members of the MIT NLP group for their suggestions and comments.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"This is a unique object for which we are able to define a proper probability model.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"BerkeleyLM uses states to optimistically search for longer n-gram matches first and must perform twice as many random accesses to retrieve backoff information.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"This larger corpus was kindly provided to us by United Informatics Inc., R.O.C. a set of initial estimates of the word frequencies.9 In this re-estimation procedure only the entries in the base dictionary were used: in other words, derived words not in the base dictionary and personal and foreign names were not used.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"The relevant variables are the set of token-level tags that appear before and after each instance of the ith word type; we denote these context pairs with the set {(tb, ta)} and they are contained in t(\u00e2\u0088\u0092i).","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"This is because different judges focused on different language pairs.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"In fact, during the first rounds many of the predictions of Th., g2 are zero.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"The intuition here is that the role of a discourse marker can usually be de 9 Both the corpus split and pre-processing code are avail-.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"Figure 2 shows examples of lexical expectations that were learned for both domains.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"orthographic words are thus only a starting point for further analysis and can only be regarded as a useful hint at the desired division of the sentence into words.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"While automatic measures are an invaluable tool for the day-to-day development of machine translation systems, they are only a imperfect substitute for human assessment of translation quality, or as the acronym BLEU puts it, a bilingual evaluation understudy.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"If e < b then the key is not found.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"36.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"While we cannot prove there are no such useful features on which one should condition trust, we can give some insight into why the features we explored offered no gain.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"We can now add a new weak hypothesis 14 based on a feature in X1 with a confidence value al hl and atl are chosen to minimize the function We now define, for 1 <i <n, the following virtual distribution, As before, Ztl is a normalization constant.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"Thus, any language that is letter equivalent to a semilinear language is also semilinear.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"The perplexity for the trigram language model used is 26:5.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"The ATB is disadvantaged by having fewer trees with longer average 5 LDC A-E catalog numbers: LDC2008E61 (ATBp1v4), LDC2008E62 (ATBp2v3), and LDC2008E22 (ATBp3v3.1).","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"The normalized judgement per sentence is the raw judgement plus (0 minus average raw judgement for this judge on this sentence).","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"The English data comes from the WSJ portion of the Penn Treebank and the other languages from the training set of the CoNLL-X multilingual dependency parsing shared task.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"For example, if a token starts with a capital letter and ends with a period (such as Mr.), then the feature InitCapPeriod is set to 1, etc. First Word: This feature group contains only one feature firstword.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"The portions of information in the large window can be individually clicked visible or invisible; here we have chosen to see (from top to bottom) \u00e2\u0080\u00a2 the full text, \u00e2\u0080\u00a2 the annotation values for the activated annotation set (co-reference), \u00e2\u0080\u00a2 the actual annotation tiers, and \u00e2\u0080\u00a2 the portion of text currently \u00e2\u0080\u0098in focus\u00e2\u0080\u0099 (which also appears underlined in the full text).","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"4.","label":0},{"source_text":"Here we present two algorithms.","target_text":"The 2(Yarowsky 95) describes the use of more sophisticated smoothing methods.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"We have not yet tried this.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"Reflexive pronouns with only 1 NP in scope..","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"am 11.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"Our work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"These systems rely on a training corpus that has been manually annotated with coreference links.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"This is not unreasonable given the application to phrase pairs from OUT, but it suggests that an interesting alternative might be to use a plain log-linear weighting function exp(Ei Aifi(s, t)), with outputs in [0, oo].","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"We identified three ways that contextual roles can be exploited: (1) by identifying caseframes that co-occur in resolutions, (2) by identifying nouns that co-occur with case- frames and using them to crosscheck anaphor\/candidate compatibility, (3) by identifying semantic classes that co- occur with caseframes and using them to crosscheck anaphor\/candidate compatability.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"The final estimating equation is then: (3) Since the total of all these class estimates was about 10% off from the Turing estimate n1\/N for the probability of all unseen hanzi, we renormalized the estimates so that they would sum to n 1jN.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"Gazdar (1985) considers a number of linguistic analyses which IG's (but not CFG's) can make, for example, the Norwedish example shown in Figure 1.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Lazy mapping reduces memory requirements by loading pages from disk only as necessary.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"Local features are features that are based on neighboring tokens, as well as the token itself.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"This means that the PCC cannot grow particularly quickly.","label":0},{"source_text":"There is no global pruning.","target_text":"The restriction can be expressed in terms of the number of uncovered source sentence positions to the left of the rightmost position m in the coverage set.","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"Assigning rhetorical relations thus poses questions that can often be answered only subjectively.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"4 53.7 43.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"TRIE uses less memory and has better locality.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"The present proposal falls into the last group.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"1 74.5 56.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"5 70.1 58.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"The final score is obtained from: max e;e0 j2fJ\u00f4\u0080\u0080\u0080L;;Jg p($je; e0) Qe0 (e; I; f1; ; Jg; j); where p($je; e0) denotes the trigram language model, which predicts the sentence boundary $ at the end of the target sentence.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"1 is given in Fig.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"This FSA I can be segmented into words by composing Id(I) with D*, to form the WFST shown in Figure 2(c), then selecting the best path through this WFST to produce the WFST in Figure 2(d).","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Nodes in the trie are based on arrays sorted by vocabulary identifier.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Making the reasonable assumption that similar information is relevant for solving these problems in Chinese, it follows that a prerequisite for intonation-boundary assignment and prominence assignment is word segmentation.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Its correct antecedent is \u00e2\u0080\u009ca revolver\u00e2\u0080\u009d, which is extracted by the caseframe \u00e2\u0080\u009ckilled with <NP>\u00e2\u0080\u009d.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"The attachment in such cases encompasses a long distance dependency that cannot be captured by Markovian processes that are typically used for morphological disambiguation.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"There are two weaknesses in Chang et al.'s model, which we improve upon.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"2.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"including Third Tone Sandhi (Shih 1986), which changes a 3 (low) tone into a 2 (rising) tone before another 3 tone: 'j\";gil, xiao3 [lao3 shu3] 'little rat,' becomes xiao3 { lao2shu3 ], rather than xiao2 { lao2shu3 ], because the rule first applies within the word lao3shu3 'rat,' blocking its phrasal application.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"The predominate focus of building systems that translate into English has ignored so far the difficult issues of generating rich morphology which may not be determined solely by local context.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"KenLM: Faster and Smaller Language Model Queries","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"At the phrasal level, we remove all function tags and traces.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"Past work however, has typically associ n = n P (Ti)P (Wi|Ti) = i=1 1 n K n ated these features with token occurrences, typically in an HMM.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"Each knowledge source then assigns a probability estimate to each candidate, which represents its belief that the candidate is the antecedent for the anaphor.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"Inflectional features marking pronominal elements may be attached to different kinds of categories marking their pronominal complements.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"This work was supported in part under the GALE program of the Defense Advanced Research Projects Agency, Contract No.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"Also, we don\u00e2\u0080\u0099t know how many such paraphrase sets are necessary to cover even some everyday things or events.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"3.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"Applications The discovered paraphrases have multiple applications.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"We checked whether the discovered links are listed in WordNet.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Reading the following record\u2019s offset indicates where the block ends.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"For the manual scoring, we can distinguish only half of the systems, both in terms of fluency and adequacy.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"63 95.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"4 7 . 3 8 . 9 2 8 . 8 2 0 . 7 3 2 . 3 3 5 . 2 2 9 . 6 2 7 . 6 1 4 . 2 4 2 . 8 4 5 . 9 4 4 . 3 6 0 . 6 6 1 . 5 4 9 . 9 3 3 . 9 Table 6: Type-level Results: Each cell report the type- level accuracy computed against the most frequent tag of each word type.","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"The entire treebank is used in the experiment, but only primary dependencies are considered.4 In all experiments, punctuation tokens are included in the data but omitted in evaluation scores.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Given the closeness of most systems and the wide over-lapping confidence intervals it is hard to make strong statements about the correlation between human judgements and automatic scoring methods such as BLEU.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"For instance, for TTS it is necessary to know that a particular sequence of hanzi is of a particular category because that knowl\u00c2\u00ad edge could affect the pronunciation; consider, for example the issues surrounding the pronunciation of ganl I qian2 discussed in Section 1.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"This is demonstrated by average scores over all systems, in terms of BLEU, fluency and adequacy, as displayed in Figure 5.","label":0},{"source_text":"Here we present two algorithms.","target_text":"With each iteration more examples are assigned labels by both classifiers, while a high level of agreement (> 94%) is maintained between them.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"We use w erations of sampling (see Figure 2 for a depiction).","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"This is summarized in Equation 5.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"If is one of Monday, Tuesday, . . .","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"We examine both the complexity of the paths of trees in the tree sets, and the kinds of dependencies that the formalisms can impose between paths.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"To make the projection practical, we rely on the twelve universal part-of-speech tags of Petrov et al. (2011).","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"JI!","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"This highly effective approach is not directly applicable to the multinomial models used for core SMT components, which have no natural method for combining split features, so we rely on an instance-weighting approach (Jiang and Zhai, 2007) to downweight domain-specific examples in OUT.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"The first point we need to address is what type of linguistic object a hanzi repre\u00c2\u00ad sents.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"The frequency of the Company \u00e2\u0080\u0093 Company domain ranks 11th with 35,567 examples.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"All systems (except for Systran, which was not tuned to Europarl) did considerably worse on outof-domain training data.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"Again, the monotone search performs worst.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"In total, for the 2,000 NE category pairs, 5,184 keywords are found.","label":0},{"source_text":"There is no global pruning.","target_text":"In Section 3, we introduce our novel concept to word reordering and a DP-based search, which is especially suitable for the translation direction from German to English.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"Because there might be some controversy about the exact definitions of such universals, this set of coarse-grained POS categories is defined operationally, by collapsing language (or treebank) specific distinctions to a set of categories that exists across all languages.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Typically, judges initially spent about 3 minutes per sentence, but then accelerate with experience.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Following Dickinson (2005), we randomly sampled 100 variation nuclei from each corpus and evaluated each sample for the presence of an annotation error.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"As the name implies, space is O(m) and linear in the number of entries.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"Information from a sentence is sometimes insufficient to classify a name correctly.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"IRST is not threadsafe.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Unfortunately, we were unable to correctly run the IRSTLM quantized variant.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"example, in Northern Mandarin dialects there is a morpheme -r that attaches mostly to nouns, and which is phonologically incorporated into the syllable to which it attaches: thus men2+r (door+R) 'door' is realized as mer2.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"We settled on contrastive evaluations of 5 system outputs for a single test sentence.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"Fort= 1,...,T:","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"Previous work has attempted to incorporate such constraints into token-level models via heavy-handed modifications to inference procedure and objective function (e.g., posterior regularization and ILP decoding) (Grac\u00c2\u00b8a et al., 2009; Ravi and Knight, 2009).","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"Any opinions, findings, conclusions, or recommendations expressed in this paper are those of the authors, and do not necessarily reflect the views of the funding organizations.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Confidence Interval: To estimate confidence intervals for the average mean scores for the systems, we use standard significance testing.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"Ideally we\u00e2\u0080\u0099d like to know the thematic role of each extracted noun phrase, but AutoSlog does not generate thematic roles.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"In order to handle the necessary word reordering as an optimization problem within our dynamic programming approach, we describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming (Held, Karp, 1962).","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Instead of offsetting new topics with punctuation, writers of MSA in sert connectives such as \u00ef\u00bf\u00bd wa and \u00ef\u00bf\u00bd fa to link new elements to both preceding clauses and the text as a whole.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"In the named entity task, X1 might be the instance space for the spelling features, X2 might be the instance space for the contextual features.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"newspaper material, but also including kungfu fiction, Buddhist tracts, and scientific material.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Approaches differ in the algorithms used for scoring and selecting the best path, as well as in the amount of contextual information used in the scoring process.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Entries landing in the same bucket are said to collide.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"The type-level posterior term can be computed according to, P (Ti|W , T \u00e2\u0088\u0092i, \u00ce\u00b2) \u00e2\u0088\u009d Note that each round of sampling Ti variables takes time proportional to the size of the corpus, as with the standard token-level HMM.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"Again, the monotone search performs worst.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Table 6: Incremental dev set results for the manually annotated grammar (sentences of length \u00e2\u0089\u00a4 70).","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"40 75.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"Thus corresponding pseudo-labels for instances on which gj abstain are set to zero and these instances do not contribute to the objective function.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Floating point values may be stored in the trie exactly, using 31 bits for non-positive log probability and 32 bits for backoff5.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"Domain adaptation is a common concern when optimizing empirical NLP applications.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"Month Names, Days of the Week, and Numbers: If is initCaps and is one of January, February, . . .","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Two annotators received training with the RST definitions and started the process with a first set of 10 texts, the results of which were intensively discussed and revised.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"This class-based model gives reasonable results: for six radical classes, Table 1 gives the estimated cost for an unseen hanzi in the class occurring as the second hanzi in a double GIVEN name.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"The first unsupervised algorithm we describe is based on the decision list method from (Yarowsky 95).","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"Pairwise comparison is done using the sign test.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"An example of a fairly low-level relation is the affix relation, which holds between a stem morpheme and an affix morpheme, such as f1 -menD (PL).","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Of course, we.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"Figure 1 depicts the lattice for a 2-words sentence bclm hneim.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"Any opinions, findings, conclusions, or recommendations expressed in this paper are those of the authors, and do not necessarily reflect the views of the funding organizations.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"Kollege.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"ni2ya3 and @5:2 xilya3, respectively.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Assuming unseen objects within each class are equiprobable, their probabilities are given by the Good-Turing theorem as: cis E( n'J.ls) Po oc N * E(N8ls) (2) where p815 is the probability of one unseen hanzi in class cls, E(n'J.15 ) is the expected number of hanzi in cls seen once, N is the total number of hanzi, and E(N(\/ 5 ) is the expected number of unseen hanzi in class cls.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"The model was parameterized such that the joint probability of a (label, feature-set) pair P(yi, xi) is written as The model assumes that (y, x) pairs are generated by an underlying process where the label is first chosen with some prior probability P(yi); the number of features mi is then chosen with some probability P(mi); finally the features are independently generated with probabilities P(xulyi).","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"It is well-known that constituency parsing models designed for English often do not generalize easily to other languages and treebanks.1 Explanations for this phenomenon have included the relative informativeness of lexicalization (Dubey and Keller, 2003; Arun and Keller, 2005), insensitivity to morphology (Cowan and Collins, 2005; Tsarfaty and Sima\u00e2\u0080\u0099an, 2008), and the effect of variable word order (Collins et al., 1999).","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"2.5 Connectives with scopes.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"able at http:\/\/nlp.stanford.edu\/projects\/arabic.shtml.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"However, lazy mapping is generally slow because queries against uncached pages must wait for the disk.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"4 65.9 48.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Given n-gram counts {cn}Nn=1, we use Flog2 c1] bits per vocabulary identifier and Flog2 cn] per index into the table of ngrams.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"Other work includes transferring latent topic distributions from source to target language for LM adaptation, (Tam et al., 2007) and adapting features at the sentence level to different categories of sentence (Finch and Sumita, 2008).","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Entries landing in the same bucket are said to collide.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"Keywords with more than one word In the evaluation, we explained that \u00e2\u0080\u009cchairman\u00e2\u0080\u009d and \u00e2\u0080\u009cvice chairman\u00e2\u0080\u009d are considered paraphrases.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"Ablation Analysis We evaluate the impact of incorporating various linguistic features into our model in Table 3.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"87 Table 7: Test set results.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"The trie data structure is commonly used for language modeling.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"In order to create an IE system for a new domain, one has to spend a long time to create the knowledge.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"AdaBoost is given access to a weak learning algorithm, which accepts as input the training examples, along with a distribution over the instances.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Keys to the table are hashed, using for example Austin Appleby\u2019s MurmurHash2, to integers evenly distributed over a large range.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"The approach recursively evaluates a quantity Q(C; j), where C is the set of already visited cities and sj is the last visited city.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Given n-gram counts {cn}Nn=1, we use Flog2 c1] bits per vocabulary identifier and Flog2 cn] per index into the table of ngrams.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"In the cases where isolated constituent precision is larger than 0.5 the affected portion of the hypotheses is negligible.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Nicola Bertoldi and Marcello Federico assisted with IRSTLM.","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"However, the accuracy is considerably higher than previously reported results for robust non-projective parsing of Czech, with a best performance of 73% UAS (Holan, 2004).","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"2.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"For corporate suffixes, a list of tokens cslist that occur frequently as the last token of an organization name is collected from the training data.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"We are especially grateful to Taylor Berg- Kirkpatrick for running additional experiments.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"MITLM 0.4 (Hsu and Glass, 2008) is mostly designed for accurate model estimation, but can also compute perplexity.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"In the label propagation stage, we propagate the automatic English tags to the aligned Italian trigram types, followed by further propagation solely among the Italian vertices. the Italian vertices are connected to an automatically labeled English vertex.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":":zhong1 : 0.0 tjl :huo2 :0.0 (R:spub:\/ic of Ch:ina) + .,_,...I : jlong4 :0.0 (mUifaty genG181) 0 \u00c2\u00a3: _NC: 40.0 Figure 3 Partial Chinese Lexicon (NC = noun; NP = proper noun).c=- - I \u00e2\u0080\u00a2=- :il: .;ss:;zhangt \u00e2\u0080\u00a2 '-:.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"If it is made up of all capital letters, then (allCaps, zone) is set to 1.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"They are: 5We are grateful to an anonymous reviewer for pointing this out.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"The other half was replaced by other participants, so we ended up with roughly the same number.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"Due to the dramatic fiscal situation in Brandenburg she now surprisingly withdrew legislation drafted more than a year ago, and suggested to decide on it not before 2003.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"What is important and is not expressed by the notation is the so-called coverage constraint: each source position j should be 'hit' exactly once by the path of the inverted alignment bI 1 = b1:::bi:::bI . Using the inverted alignments in the maximum approximation, we obtain as search criterion: max I (p(JjI) max eI 1 ( I Yi=1 p(eijei\u00f4\u0080\u0080\u00801 i\u00f4\u0080\u0080\u00802) max bI 1 I Yi=1 [p(bijbi\u00f4\u0080\u0080\u00801; I; J) p(fbi jei)])) = = max I (p(JjI) max eI 1;bI 1 ( I Yi=1 p(eijei\u00f4\u0080\u0080\u00801 i\u00f4\u0080\u0080\u00802) p(bijbi\u00f4\u0080\u0080\u00801; I; J) p(fbi jei)])); where the two products over i have been merged into a single product over i. p(eijei\u00f4\u0080\u0080\u00801 i\u00f4\u0080\u0080\u00802) is the trigram language model probability.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"Naseem et al. (2009) and Snyder et al.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"64 94.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"Annotators argued for the importance of having correct and even multiple references.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"However, it is unlikely that other occurrences of News Broadcasting Corp. in the same document also co-occur with Even.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"The ability to redistribute belief values across sets rather than individual hypotheses is key.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Our code is thread-safe, and integrated into the Moses, cdec, and Joshua translation systems.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"37 84.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"We utilized two kinds of datasets in our experiments: (i) monolingual treebanks9 and (ii) large amounts of parallel text with English on one side.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"The work of the second author as well as collaboration visits to Israel was financed by NWO, grant number 017.001.271.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"Dictionaries: Due to the limited amount of training material, name dictionaries have been found to be useful in the named entity task.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"This is in general very difficult, given the extremely free manner in which Chinese given names are formed, and given that in these cases we lack even a family name to give the model confidence that it is identifying a name.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"The predominate focus of building systems that translate into English has ignored so far the difficult issues of generating rich morphology which may not be determined solely by local context.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"0 . 8 3 1 0.859 496 76.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"2.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"(Brandt 1996) extended these ideas toward a conception of kommunikative Gewichtung (\u00e2\u0080\u0098communicative-weight assignment\u00e2\u0080\u0099).","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Typically, judges initially spent about 3 minutes per sentence, but then accelerate with experience.","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"Thus, the arc from je to jedna will be labeled 5b\u2193 (to indicate that there is a syntactic head below it).","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"Oracle results).","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"We have already mentioned the closely related work by Matsoukas et al (2009) on discriminative corpus weighting, and Jiang and Zhai (2007) on (nondiscriminative) instance weighting.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"92 77.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"We must adjoin all trees in an auxiliary tree set together as a single step in the derivation.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"Each out-of-domain phrase pair is characterized by a set of simple features intended to reflect how useful it will be.","label":1},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"For the automatic evaluation, we used BLEU, since it is the most established metric in the field.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Space- or punctuation-delimited * 700 Mountain Avenue, 2d451, Murray Hill, NJ 07974, USA.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"For the disasters domain, 8245 texts were used for training and the 40 test documents contained 447 anaphoric links.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"The perplexity for the trigram language model used is 26:5.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Because of this, we retokenized and lowercased submitted output with our own tokenizer, which was also used to prepare the training and test data.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"It was our hope that this competition, which included the manual and automatic evaluation of statistical systems and one rulebased commercial system, will give further insight into the relation between automatic and manual evaluation.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Because b is a function, no additional hypothesis splitting happens.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"However, lexically similar NPs usually refer to the same entity in two cases: proper names and existential noun phrases.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"\u00e2\u0080\u0094 similar results have been observed across multiple languages.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Under this scheme, n human judges are asked independently to segment a text.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Ex: He was found in San Jose, where ...","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"immediately by a Romanization into the pinyin transliteration scheme; numerals following each pinyin syllable represent tones.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"Since the early days of statistical NLP, researchers have observed that a part-of-speech tag distribution exhibits \u00e2\u0080\u009cone tag per discourse\u00e2\u0080\u009d sparsity \u00e2\u0080\u0094 words are likely to select a single predominant tag in a corpus, even when several tags are possible.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"Presenting the output of several system allows the human judge to make more informed judgements, contrasting the quality of the different systems.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"The estimation of the probabilities in the model is carried out as shown in Equation 4.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"In Section 2, we brie y review our approach to statistical machine translation.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Not every annotator was fluent in both the source and the target language.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"Link phrases based on instance pairs Using NE instance pairs as a clue, we find links between sets of phrases.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"However, ince we extracted the test corpus automatically from web sources, the reference translation was not always accurate \u2014 due to sentence alignment errors, or because translators did not adhere to a strict sentence-by-sentence translation (say, using pronouns when referring to entities mentioned in the previous sentence).","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"The transition from f, to a final state transduces c to the grammatical tag PL with cost cost(unseen(f,)): cost(i\u00c2\u00a5JJ1l.ir,) == cost(i\u00c2\u00a5JJ1l.)","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"Table 5 provides insight into the behavior of different models in terms of the tagging lexicon they generate.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"The discovered paraphrases can be a big help to reduce human labor and create a more comprehensive pattern set.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"The anaphor and antecedent appear in boldface.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"However, there are several reasons why this approach will not in general work: 1.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"For this reason, na\u00efve Bayes classifiers are well-matched to this problem.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"The second setting uses the news-related subcorpora for the NIST09 MT Chinese to English evaluation8 as IN, and the remaining NIST parallel Chinese\/English corpora (UN, Hong Kong Laws, and Hong Kong Hansard) as OUT.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"A moment's reflection will reveal that things are not quite that simple.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Memory usage in PROBING is high, though SRILM is even larger, so where memory is of concern we recommend using TRIE, if it fits in memory.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"\u00e2\u0080\u00a2 Some tools would allow for the desired annotation mode, but are so complicated (they can be used for many other purposes as well) that annotators take a long time getting used to them.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"For a corpus C, let M be the set of tuples \u00e2\u0088\u0097n, l), where n is an n-gram with bracketing label l. If any n appears 6 Generative parsing performance is known to deteriorate with sentence length.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"We computed BLEU scores for each submission with a single reference translation.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"By establishing significantly higher parsing baselines, we have shown that Arabic parsing performance is not as poor as previously thought, but remains much lower than English.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"The Stanford parser includes both the manually annotated grammar (\u00c2\u00a74) and an Arabic unknown word model with the following lexical features: 1.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"For the disasters domain, 8245 texts were used for training and the 40 test documents contained 447 anaphoric links.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"Our method does not assume any knowledge about the target language (in particular no tagging dictionary is assumed), making it applicable to a wide array of resource-poor languages.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"We substantially outperform all of them on query speed and offer lower memory consumption than lossless alternatives.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"For instance, the sentence Similar improvements in haemoglobin levels were reported in the scientific literature for other epoetins would likely be considered domain-specific despite the presence of general phrases like were reported in.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Out-of-Vocabulary: We derived a lexicon list from WordNet 1.6, and words that are not found in this list have a feature out-of-vocabulary set to 1.","label":0},{"source_text":"There is no global pruning.","target_text":"input: source string f1:::fj :::fJ initialization for each cardinality c = 1; 2; ; J do for each pair (C; j), where j 2 C and jCj = c do for each target word e 2 E Qe0 (e; C; j) = p(fj je) max \u00c3\u0086;e00 j02Cnfjg fp(jjj0; J) p(\u00c3\u0086) p\u00c3\u0086(eje0; e00) Qe00 (e0;C n fjg; j0)g words fj in the input string of length J. For the final translation each source position is considered exactly once.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Since we could not bias the subjects towards a particular segmentation and did not presume linguistic sophistication on their part, the instructions were simple: subjects were to mark all places they might plausibly pause if they were reading the text aloud.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"0.2 0.1 0.0 -0.1 25 26 27 28 29 30 31 32 -0.2 -0.3 \u2022systran \u2022 ntt 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 20 21 22 23 24 25 26 Fluency Fluency \u2022systran \u2022nrc rali 25 26 27 28 29 30 31 32 0.2 0.1 0.0 -0.1 -0.2 -0.3 cme p \ufffd 20 21 22 23 24 25 26 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 Figure 14: Correlation between manual and automatic scores for English-French 119 In Domain Out of Domain \u2022upv Adequacy -0.9 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 \u2022upv 23 24 25 26 27 28 29 30 31 32 \u2022upc-mr \u2022utd \u2022upc-jmc \u2022uedin-birch \u2022ntt \u2022rali \u2022uedin-birch 16 17 18 19 20 21 22 23 24 25 26 27 Adequacy \u2022upc-mr 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 -0.9 -1.0 -1.1 English-Spanish Fluency \u2022ntt \u2022nrc \u2022rali \u2022uedin-birch -0.2 -0.3 -0.5 \u2022upv 16 17 18 19 20 21 22 23 24 25 26 27 -0.4 nr \u2022 rali Fluency -0.4 \u2022upc-mr utd \u2022upc-jmc -0.5 -0.6 \u2022upv 23 24 25 26 27 28 29 30 31 32 0.2 0.1 -0.0 -0.1 -0.2 -0.3 0.3 0.2 0.1 -0.0 -0.1 -0.6 -0.7 Figure 15: Correlation between manual and automatic scores for English-Spanish 120 English-German In Domain Out of Domain Adequacy Adequacy 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 \u2022upv -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 -0.9 \u2022upv 0.5 0.4 \u2022systran \u2022upc-mr \u2022 \u2022rali 0.3 \u2022ntt 0.2 0.1 -0.0 -0.1 \u2022systran \u2022upc-mr -0.9 9 10 11 12 13 14 15 16 17 18 19 6 7 8 9 10 11 Fluency 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 \u2022upv -0.5 \u2022upv \u2022systran \u2022upc-mr \u2022 Fluency 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 \u2022systran \u2022ntt","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"A high-level relation is agent, which relates an animate nominal to a predicate.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"Section 2 describes our baseline techniques for SMT adaptation, and section 3 describes the instance-weighting approach.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Borth- wick (1999) successfully made use of other hand- coded systems as input for his MENE system, and achieved excellent results.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"It is the performance we could achieve if an omniscient observer told us which parser to pick for each of the sentences.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"f;g denotes the empty set, where no source sentence position is covered.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Thus in a two-hanzi word like lflli?J zhong1guo2 (middle country) 'China' there are two syllables, and at the same time two morphemes.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"Given parameter estimates, the label for a test example x is defined as We should note that the model in equation 9 is deficient, in that it assigns greater than zero probability to some feature combinations that are impossible.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"The transition from f, to a final state transduces c to the grammatical tag PL with cost cost(unseen(f,)): cost(i\u00c2\u00a5JJ1l.ir,) == cost(i\u00c2\u00a5JJ1l.)","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"Our clue is the NE instance pairs.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"However, it is also mostly political content (even if not focused on the internal workings of the European Union) and opinion.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"In addition to a heuristic based on decision list learning, we also presented a boosting-like framework that builds on ideas from (Blum and Mitchell 98).","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"We include a constituent in our hypothesized parse if it appears in the output of a majority of the parsers.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Otherwise, it is set to 0.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"It did 402 queries\/ms using 1.80 GB. cMemory use increased during scoring due to batch processing (MIT) or caching (Rand).","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"As each global feature group is added to the list of features, we see improvements to both MUC6 and MUC6 MUC7 Baseline 90.75% 85.22% + ICOC 91.50% 86.24% + CSPP 92.89% 86.96% + ACRO 93.04% 86.99% + SOIC 93.25% 87.22% + UNIQ 93.27% 87.24% Table 3: F-measure after successive addition of each global feature group Table 5: Comparison of results for MUC6 Systems MUC6 MUC7 No.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"This offers the well-known advantages for inter- changability, but it raises the question of how to query the corpus across levels of annotation.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"Cohen and Smith (2007) followed up on these results and proposed a system for joint inference of morphological and syntactic structures using factored models each designed and trained on its own.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"2 56.2 32.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"set was based on an earlier version of the Chang et a!.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"Having explained the various layers of annotation in PCC, we now turn to the question what all this might be good for.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"The counts represent portions of the approximately 44000 constituents hypothesized by the parsers in the development set.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"In general, the lemma of the previous section does not ensure that all the productions in the combined parse are found in the grammars of the member parsers.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"In this way we restrict the parameterization of a Language Original case English Danish Dutch German Spanish Swedish Portuguese 94.6 96.3 96.6 95.5 95.4 93.3 95.6 Table 1: Upper bound on tagging accuracy assuming each word type is assigned to majority POS tag.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Our baseline for all sentence lengths is 5.23% F1 higher than the best previous result.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"The goal of our research was to explore the use of contextual role knowledge for coreference resolution.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"One side of the decision making process is when we choose to believe a constituent should be in the parse, even though only one parser suggests it.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"A total of 13,976 phrases were grouped.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"They are: 5We are grateful to an anonymous reviewer for pointing this out.","label":0},{"source_text":"Here we present two algorithms.","target_text":"We then discuss how we adapt and generalize a boosting algorithm, AdaBoost, to the problem of named entity classification.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"Previous work has attempted to incorporate such constraints into token-level models via heavy-handed modifications to inference procedure and objective function (e.g., posterior regularization and ILP decoding) (Grac\u00c2\u00b8a et al., 2009; Ravi and Knight, 2009).","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"The supervised POS tagging accuracies (on this tagset) are shown in the last row of Table 2.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"The improvement is due to the cost of bit-level reads and avoiding reads that may fall in different virtual memory pages.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"(2009).","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"Since every CFL is known to be semilinear (Parikh, 1966), in order to show semilinearity of some language, we need only show the existence of a letter equivalent CFL Our definition of LCFRS's insists that the composition operations are linear and nonerasing.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"In this way we restrict the parameterization of a Language Original case English Danish Dutch German Spanish Swedish Portuguese 94.6 96.3 96.6 95.5 95.4 93.3 95.6 Table 1: Upper bound on tagging accuracy assuming each word type is assigned to majority POS tag.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"If we remove this sample from the evaluation, then the ATB type-level error rises to only 37.4% while the n-gram error rate increases to 6.24%.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"In all cases, the key is collapsed to its 64-bit hash.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"10.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"In most cases, however, these expansions come with a steep increase in model complexity, with respect to training procedure and inference time.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"2 70.7 52.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"Feature selection is implemented using a feature cutoff: features seen less than a small count during training will not be used.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"We present two algorithms.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"In addition to a heuristic based on decision list learning, we also presented a boosting-like framework that builds on ideas from (Blum and Mitchell 98).","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"6 Results and Analysis.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"Two processes are spawned requiring B to derive z,.., and C to derive yi , , y,.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Affix Pron Base category N found N missed (recall) N correct (precision) t,-,7 The second issue is that rare family names can be responsible for overgeneration, especially if these names are otherwise common as single-hanzi words.","label":0},{"source_text":"Here we present two algorithms.","target_text":"On each step CoBoost searches for a feature and a weight so as to minimize either 40 or 40.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"So, who won the competition?","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"Semilinearity and the closely related constant growth property (a consequence of semilinearity) have been discussed in the context of grammars for natural languages by Joshi (1983\/85) and Berwick and Weinberg (1984).","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"On each step CoBoost searches for a feature and a weight so as to minimize either 40 or 40.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"While possible to utilize the feature-based log-linear approach described in Berg-Kirkpatrick et al.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"We define the following function: If Zco is small, then it follows that the two classifiers must have a low error rate on the labeled examples, and that they also must give the same label on a large number of unlabeled instances.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"(1992).","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"These estimates are in turn combined linearly with relative-frequency estimates from an in-domain phrase table.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"E.g. when 'Zahnarzttermin' is aligned to dentist's, the extended lexicon model might learn that 'Zahnarzttermin' actuallyhas to be aligned to both dentist's and ap pointment.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"To check for this, we do pairwise bootstrap resampling: Again, we repeatedly sample sets of sentences, this time from both systems, and compare their BLEU scores on these sets.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"We have estimated the performance of IdentiFinder ' 99 at 200K words of training data from the graphs.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"It is not immediately obvious how to formulate an equivalent to equation (1) for an adapted TM, because there is no well-defined objective for learning TMs from parallel corpora.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"A k-tape ATM, M, has a read-only input tape and k read-write work tapes.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"(Again, the goal of also in structural features.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":", December, then the feature MonthName is set to 1.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":".., Tn ) T W \u00cf\u0086 E w : Token word seqs (obs) t : Token tag assigns (det by T ) PARAMETERS \u00cf\u0088 : Lexicon parameters \u00ce\u00b8 : Token word emission parameters \u00cf\u0086 : Token tag transition parameters \u00cf\u0086 \u00cf\u0086 t1 t2 \u00ce\u00b8 \u00ce\u00b8 w1 w2 K \u00cf\u0086 T tm O K \u00ce\u00b8 E wN m N N Figure 1: Graphical depiction of our model and summary of latent variables and parameters.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"One annotator suggested that this was the case for as much as 10% of our test sentences.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"An initial step of any text\u00c2\u00ad analysis task is the tokenization of the input into words.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"Pairwise comparison: We can use the same method to assess the statistical significance of one system outperforming another.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"Other work includes transferring latent topic distributions from source to target language for LM adaptation, (Tam et al., 2007) and adapting features at the sentence level to different categories of sentence (Finch and Sumita, 2008).","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Second, comparisons of different methods are not meaningful unless one can eval\u00c2\u00ad uate them on the same corpus.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"One implementation issue deserves some elaboration.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"We have presented a method for unsupervised part- of-speech tagging that considers a word type and its allowed POS tags as a primary element of the model.","label":0},{"source_text":"Here we present two algorithms.","target_text":"The following algorithm was then used to induce new rules: Let Count' (x) be the number of times feature x is seen with some known label in the training data.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"The F- measure score increased for both domains, reflecting a substantial increase in recall with a small decrease in precision.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"Step 2.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"\u00e2\u0080\u00a2 Anaphoric links: the annotator is asked to specify whether the anaphor is a repetition, partial repetition, pronoun, epithet (e.g., Andy Warhol \u00e2\u0080\u0093 the PopArt artist), or is-a (e.g., Andy Warhol was often hunted by photographers.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"MITLM 0.4 (Hsu and Glass, 2008) is mostly designed for accurate model estimation, but can also compute perplexity.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"We received submissions from 14 groups from 11 institutions, as listed in Figure 2.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"2) An improved language model, which takes into account syntactic structure, e.g. to ensure that a proper English verbgroup is generated.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"The entries in this table can be compared with those of Table 3 to see how the performance of the combining techniques degrades in the presence of an inferior parser.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"As can be seen, GR and this \"pared-down\" statistical method perform quite similarly, though the statistical method is still slightly better.16 AG clearly performs much less like humans than these methods, whereas the full statistical algorithm, including morphological derivatives and names, performs most closely to humans among the automatic methods.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Overall, language modeling significantly impacts decoder performance.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"\u2022 We evaluated translation from English, in addition to into English.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"In our model, we associate these features at the type-level in the lexicon.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"The complexity of the quasimonotone search is O(E3 J (R2+LR)).","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"Hence we decided to restrict ourselves to only information from the same document.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"To be short, we omit the target words e; e0 in the formulation of the search hypotheses.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"It may seem surprising to some readers that the interhuman agreement scores reported here are so low.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"Without HSPELL-pruning, our simpler grammars are somewhat lagging behind, but as the grammars improve the gap is bridged.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Many researchers have developed coreference resolvers, so we will only discuss the methods that are most closely related to BABAR.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"The high 1 tone of J1l would not normally neutralize in this fashion if it were functioning as a word on its own.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"The matching sentence pairs are then added to the IN corpus, and the system is re-trained.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"phrase (markContainsVerb).","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"By this assumption, each element x E X can also be represented as (xi, x2) E X1 x X2.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"In the initial release of the ATB, inter-annotator agreement was inferior to other LDC treebanks (Maamouri et al., 2008).","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"The size of TRIE is particularly sensitive to F1092 c11, so vocabulary filtering is quite effective at reducing model size.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"Fall.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"Given that weights on all outgoing arcs sum up to one, weights induce a probability distribution on the lattice paths.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"In this paper we have proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"02 99.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"The authors provided us with a ratio between TPT and SRI under different conditions. aLossy compression with the same weights. bLossy compression with retuned weights. ditions make the value appropriate for estimating repeated run times, such as in parameter tuning.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"The judgement of 4 in the first case will go to a vastly better system output than in the second case.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"In the appositive case, the contextual predictor was the head of the modifying appositive (president in the Maury Cooper example); in the second case, the contextual predictor was the preposition together with the noun it modifies (plant_in in the Georgia example).","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"(7) is at 0 when: 1) Vi : sign(gi (xi)) = sign(g2 (xi)); 2) Ig3(xi)l oo; and 3) sign(gi (xi)) = yi for i = 1, , m. In fact, Zco provides a bound on the sum of the classification error of the labeled examples and the number of disagreements between the two classifiers on the unlabeled examples.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"The returned state s(wn1) may then be used in a followon query p(wn+1js(wn1)) that extends the previous query by one word.","label":0},{"source_text":"There is no global pruning.","target_text":"4.3 Translation Experiments.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"It can also be seen clearly in this plot that two of the Taiwan speakers cluster very closely together, and the third Tai\u00c2\u00ad wan speaker is also close in the most significant dimension (the x axis).","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"So, there is a limitation that IE can only be performed for a predefined task, like \u00e2\u0080\u009ccorporate mergers\u00e2\u0080\u009d or \u00e2\u0080\u009cmanagement succession\u00e2\u0080\u009d.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Moses keeps language models and many other resources in static variables, so these are still resident in memory.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"Moving beyond directly related work, major themes in SMT adaptation include the IR (Hildebrand et al., 2005; L\u00a8u et al., 2007; Zhao et al., 2004) and mixture (Finch and Sumita, 2008; Foster and Kuhn, 2007; Koehn and Schroeder, 2007; L\u00a8u et al., 2007) approaches for LMs and TMs described above, as well as methods for exploiting monolingual in-domain text, typically by translating it automatically and then performing self training (Bertoldi and Federico, 2009; Ueffing et al., 2007; Schwenk and Senellart, 2009).","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"However, it is also mostly political content (even if not focused on the internal workings of the European Union) and opinion.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"The main disadvantage of manual evaluation is that it is time-consuming and thus too expensive to do frequently.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"The transitive closure of the dictionary in (a) is composed with Id(input) (b) to form the WFST (c).","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"Annotators suggested that long sentences are almost impossible to judge.","label":0},{"source_text":"Here we present two algorithms.","target_text":"I = 1X21 N and N is a &quot;medium&quot; sized number so that it is feasible to collect 0(N) unlabeled examples.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"The goal of machine translation is the translation of a text given in some source language into a target language.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"It can be shown that this objective is convex in q.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"Table 5: Individual Performance of KSs for Disasters (e.g., \u00e2\u0080\u009cthe mayor\u00e2\u0080\u009d vs. \u00e2\u0080\u009cthe journalist\u00e2\u0080\u009d).","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"t(i).","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"irL as the product of the probability estimate for i\u00c2\u00a5JJ1l., and the probability estimate just derived for unseen plurals in ir,: p(i\u00c2\u00a51J1l.ir,) p(i\u00c2\u00a51J1l.)p(unseen(f,)).","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"In contrast to the Bayesian HMM, \u00ce\u00b8t is not drawn from a distribution which has support for each of the n word types.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"MENE has only been tested on MUC7.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"It was our hope that this competition, which included the manual and automatic evaluation of statistical systems and one rulebased commercial system, will give further insight into the relation between automatic and manual evaluation.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"The difference in performance between pronouns and definite noun phrases surprised us.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"4.3 Translation Experiments.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"They considered a semi-supervised POS tagging scenario and showed that one can use a graph over trigram types, and edge weights based on distributional similarity, to improve a supervised conditional random field tagger.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"This section measures performance on shared tasks in order of increasing complexity: sparse lookups, evaluating perplexity of a large file, and translation with Moses.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"Second, the reduced number of hidden variables and parameters dramatically speeds up learning and inference.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"A spelling rule might be a simple look-up for the string (e.g., a rule that Honduras is a location) or a rule that looks at words within a string (e.g., a rule that any string containing Mr. is a person).","label":0},{"source_text":"Here we present two algorithms.","target_text":"Thus the method makes the fairly strong assumption that the features can be partitioned into two types such that each type alone is sufficient for classification.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"We have presented a method for unsupervised part- of-speech tagging that considers a word type and its allowed POS tags as a primary element of the model.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"For the TM, this is: where cI(s, t) is the count in the IN phrase table of pair (s, t), po(s|t) is its probability under the OUT TM, and cI(t) = &quot;s, cI(s', t).","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"Table 3: Training and test conditions for the Verbmobil task (*number of words without punctuation marks).","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"The development of automatic scoring methods is an open field of research.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"If it is made up of all capital letters, then (allCaps, zone) is set to 1.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Given a document to process, BABAR uses four modules to perform coreference resolution.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"of Articles No.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Also, the argument has been made that machine translation performance should be evaluated via task-based evaluation metrics, i.e. how much it assists performing a useful task, such as supporting human translators or aiding the analysis of texts.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"For this reason, na\u00efve Bayes classifiers are well-matched to this problem.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"Learned Tag Prior (PRIOR) We next assume there exists a single prior distribution \u00cf\u0088 over tag assignments drawn from DIRICHLET(\u00ce\u00b2, K ).","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"We use w erations of sampling (see Figure 2 for a depiction).","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"The test accuracy more or less asymptotes.","label":0},{"source_text":"Here we present two algorithms.","target_text":"This procedure is repeated for T rounds while alternating between the two classifiers.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"It differs from previous machine learning-based NERs in that it uses information from the whole document to classify each word, with just one classifier.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"We also mark all tags that dominate a word with the feminine ending :: taa mar buuTa (markFeminine).","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"3 54.4 33.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"Learned Tag Prior (PRIOR) We next assume there exists a single prior distribution \u00cf\u0088 over tag assignments drawn from DIRICHLET(\u00ce\u00b2, K ).","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"It is. based on the traditional character set rather than the simplified character set used in Singapore and Mainland China.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"The first stage identifies a keyword in each phrase and joins phrases with the same keyword into sets.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"We concentrate on those sets.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"Xim, } associated with the ith example.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"This suggests that the backoff model is as reasonable a model as we can use in the absence of further information about the expected cost of a plural form.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"Under certain conditions the constituent voting and na\u00efve Bayes constituent combination techniques are guaranteed to produce sets of constituents with no crossing brackets.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"The evidence may not say anything about whether A is more likely than B, only that C and D are not likely.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"For example, if {N P1, N P2, N P3} are all coreferent, then each NP must be linked to one of the other two NPs.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"In graph-based learning approaches one constructs a graph whose vertices are labeled and unlabeled examples, and whose weighted edges encode the degree to which the examples they link have the same label (Zhu et al., 2003).","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"We can better predict the probability of an unseen hanzi occurring in a name by computing a within-class Good-Turing estimate for each radical class.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"Final (F): The rest of the sentence is processed monotonically taking account of the already covered positions.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"12 One class of full personal names that this characterization does not cover are married women's names.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"For example, hanzi containing the INSECT radical !R tend to denote insects and other crawling animals; examples include tr wal 'frog,' feng1 'wasp,' and !Itt she2 'snake.'","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"This step ensures that the most frequent terms for each domain are labeled (in case some of them are not in WordNet) and labeled with the sense most appropriate for the domain.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"Supervised part-of-speech (POS) taggers, for example, approach the level of inter-annotator agreement (Shen et al., 2007, 97.3% accuracy for English).","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"Suffixes and Prefixes: This group contains only two features: Corporate-Suffix and Person-Prefix.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"The second algorithm extends ideas from boosting algorithms, designed for supervised learning tasks, to the framework suggested by (Blum and Mitchell 98).","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"Before explaining our method in detail, we present a brief overview in this subsection.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Finally, as\u00c2\u00ad suming a simple bigram backoff model, we can derive the probability estimate for the particular unseen word i\u00c2\u00a51J1l.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"We would also like to thank Amarnag Subramanya for helping us with the implementation of label propagation and Shankar Kumar for access to the parallel data.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Unigram records store probability, backoff, and an index in the bigram table.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"(2009).","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"On the other hand, we are interested in the application of rhetorical analysis or \u00e2\u0080\u0098discourse parsing\u00e2\u0080\u0099 (3.2 and 3.3), in text generation (3.4), and in exploiting the corpus for the development of improved models of discourse structure (3.5).","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"We have developed a coreference resolver called BABAR that uses contextual role knowledge to make coreference decisions.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"If the token is the first word of a sentence, then this feature is set to 1.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"Hyperparameters Our model has two Dirichlet concentration hyperparameters: \u00ce\u00b1 is the shared hyperparameter for the token-level HMM emission and transition distributions.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Under this scheme, n human judges are asked independently to segment a text.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"For the inverted alignment probability p(bijbi\u00f4\u0080\u0080\u00801; I; J), we drop the dependence on the target sentence length I. 2.2 Word Joining.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Our system does not currently make use of titles, but it would be straightforward to do so within the finite-state framework that we propose.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"The task is to learn a function from an input string (proper name) to its type, which we will assume to be one of the categories Person, Organization, or Location.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"Fortunately, performance was stable across various values, and we were able to use the same hyperparameters for all languages.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"For the PROBING implementation, hash table sizes are in the millions, so the most relevant values are on the right size of the graph, where linear probing wins.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"For verbs we add two features.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"In the named entity task, X1 might be the instance space for the spelling features, X2 might be the instance space for the contextual features.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"The remaining arcs are marked OOV.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"This process is repeated 5 times by rotating the data appropriately.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"In the second scheme, Head+Path, we in addition modify the label of every arc along the lifting path from the syntactic to the linear head so that if the original label is p the new label is p\u2193.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"SRILM (Stolcke, 2002) is widely used within academia.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"We offer a state function s(wn1) = wn\ufffd where substring wn\ufffd is guaranteed to extend (to the right) in the same way that wn1 does for purposes of language modeling.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"Given a PCFG grammar G and a lattice L with nodes n1 ... nk, we construct the weighted grammar GL as follows: for every arc (lexeme) l E L from node ni to node nj, we add to GL the rule [l --+ tni, tni+1, ... , tnj_1] with a probability of 1 (this indicates the lexeme l spans from node ni to node nj).","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"At first glance, this seems only peripherally related to our work, since the specific\/general distinction is made for features rather than instances.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"Finally, we intend to explore more sophisticated instanceweighting features for capturing the degree of generality of phrase pairs.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Ex: The government said it ...","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"2.1 Part-of-speech tags.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"html 4 www.wagsoft.com\/RSTTool assigning rhetorical relations is a process loaded with ambiguity and, possibly, subjectivity.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"The Leaf Ancestor metric measures the cost of transforming guess trees to the reference (Sampson and Babarczy, 2003).","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"MITLM 0.4 (Hsu and Glass, 2008) is mostly designed for accurate model estimation, but can also compute perplexity.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"58 95.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"(In this figure eps is c) be implemented, though, such as a maximal-grouping strategy (as suggested by one reviewer of this paper); or a pairwise-grouping strategy, whereby long sequences of unattached hanzi are grouped into two-hanzi words (which may have some prosodic motivation).","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"More recently, (Riloff and Jones 99) describe a method they term &quot;mutual bootstrapping&quot; for simultaneously constructing a lexicon and contextual extraction patterns.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"For illustration, an English translation of one of the commentaries is given in Figure 1.","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"In order to facilitate this task, we extend the set of arc labels to encode information about lifting operations.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"Each distance in the traveling salesman problem now corresponds to the negative logarithm of the product of the translation, alignment and language model probabilities.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"D o m ai n # of ph ras es t o t a l p h r a s e s ac cu ra cy C C 7 o r m o r e 1 0 5 8 7 . 6 % 6 o r l e s s 1 0 6 6 7 . 0 % P C 7 o r m o r e 3 5 9 9 9 . 2 % 6 o r l e s s 2 5 5 6 5 . 1 % Table 1.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"For each 13 Of course, this weighting makes the PCFG an improper distribution.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Table 1 The cost as a novel given name (second position) for hanzi from various radical classes.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"However, some caveats are in order in comparing this method (or any method) with other approaches to seg\u00c2\u00ad mentation reported in the literature.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"This logic applies recursively: if wnf+1 similarly does not extend and has zero log backoff, it too should be omitted, terminating with a possibly empty context.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"A search restriction especially useful for the translation direction from German to English is presented.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"This means that the rules in our grammar are of two kinds: (a) syntactic rules relating nonterminals to a sequence of non-terminals and\/or PoS tags, and (b) lexical rules relating PoS tags to lattice arcs (lexemes).","label":0},{"source_text":"There is no global pruning.","target_text":"A typical situation is shown in Figure 1.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"2.3 Rhetorical structure.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"Step 4.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"Also, the argument has been made that machine translation performance should be evaluated via task-based evaluation metrics, i.e. how much it assists performing a useful task, such as supporting human translators or aiding the analysis of texts.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"We resolve this problem by inserting an entry with probability set to an otherwise-invalid value (\u2212oc).","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"This PCFG is incorporated into the Stanford Parser, a factored model that chooses a 1-best parse from the product of constituency and dependency parses.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"The P (W |T , \u00cf\u0088) term in the lexicon component now decomposes as: n P (W |T , \u00cf\u0088) = n P (Wi|Ti, \u00cf\u0088) i=1 n \u00ef\u00a3\u00ab \u00ef\u00a3\u00b6 tions are not modeled by the standard HMM, which = n \u00ef\u00a3\u00ad n P (v|\u00cf\u0088Ti f )\u00ef\u00a3\u00b8 instead can model token-level frequency.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"But gold segmentation is not available in application settings, so a segmenter and parser are arranged in a pipeline.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"If the expression is a word or a short phrase (like \u00e2\u0080\u009ccorporation\u00e2\u0080\u009d and \u00e2\u0080\u009ccompany\u00e2\u0080\u009d), it is called a \u00e2\u0080\u009csynonym\u00e2\u0080\u009d.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"There are thus some very good reasons why segmentation into words is an important task.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"\u201cin the shadow\u201d.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"It also uses less memory, with 8 bytes of overhead per entry (we store 16-byte entries with m = 1.5); linked list implementations hash set and unordered require at least 8 bytes per entry for pointers.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"). context=x The context for the entity.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"4.2 A Sample Segmentation Using Only Dictionary Words Figure 4 shows two possible paths from the lattice of possible analyses of the input sentence B X:\u00c2\u00a5 .:.S:P:l 'How do you say octopus in Japanese?' previously shown in Figure 1.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"Finally, our U (unparsed) measure is used to report the number of sentences to which our system could not propose a joint analysis.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"All annotations are done with specific tools and in XML; each layer has its own DTD.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"3.2 Results.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"It is worth noting that the middle words of the Italian trigrams are nouns too, which exhibits the fact that the similarity metric connects types having the same syntactic category.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"For that application, at a minimum, one would want to know the phonological word boundaries.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"The question of what soft function to pick, and how to design' algorithms which optimize it, is an open question, but appears to be a promising way of looking at the problem.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"Equation 2 is an estimate of the conditional probability of the label given the feature, P(yjx).","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"Recent results (e.g., (Yarowsky 95; Brill 95; Blum and Mitchell 98)) have suggested that unlabeled data can be used quite profitably in reducing the need for supervision.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Figure 2 shows examples of lexical expectations that were learned for both domains.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"In our particular case the majority requires the agreement of only two parsers because we have only three.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"The code is opensource, has minimal dependencies, and offers both C++ and Java interfaces for integration.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"Inspection of the data shows that at n = 2500, the two classifiers both give labels on 44,281 (49.2%) of the unlabeled examples, and give the same label on 99.25% of these cases.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Tokens were converted to vocabulary identifiers in advance and state was carried from each query to the next.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"Every parse \u03c0 selects a specific morphological segmentation (l1...lk) (a path through the lattice).","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"The results of this experiment can be seen in Table 5.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"The general idea for the knowledge- based part is to have the system use as much information as it can find at its disposal to produce a target representation as specific as possible and as underspecified as necessary.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"newspaper material, but also including kungfu fiction, Buddhist tracts, and scientific material.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"Microsoft\u2019s approach uses dependency trees, others use hierarchical phrase models.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"The advantage is that we can recombine search hypotheses by dynamic programming.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Note also that the costs currently used in the system are actually string costs, rather than word costs.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"If evidence indicates that hypotheses C and D are less likely than hypotheses A and B, then probabilities are redistributed to reflect the fact that {A, B} is more likely to contain the answer than {C, D}.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"The set of n-grams appearing in a model is sparse, and we want to efficiently find their associated probabilities and backoff penalties.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"Limitations of (Blum and Mitchell 98): While the assumptions of (Blum and Mitchell 98) are useful in developing both theoretical results and an intuition for the problem, the assumptions are quite limited.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"We will return to these issues in the discussion section.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"The goal of our research was to explore the use of contextual role knowledge for coreference resolution.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"For the automatic scoring method BLEU, we can distinguish three quarters of the systems.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"att.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"This paper does not necessarily reflect the position of the U.S. Government.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"The 2nd block contains the IR system, which was tuned by selecting text in multiples of the size of the EMEA training corpus, according to dev set performance.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"For previously unseen hanzi in given names, Chang et al. assign a uniform small cost; but we know that some unseen hanzi are merely acci\u00c2\u00ad dentally missing, whereas others are missing for a reason-for example, because they have a bad connotation.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"English parsing evaluations usually report results on sentences up to length 40.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"The fraction of buckets that are empty is m\u22121 m , so average lookup time is O( m 1) and, crucially, constant in the number of entries.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Evaluation of Morphological Analysis.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"These 140 NE categories are designed by extending MUC\u00e2\u0080\u0099s 7 NE categories with finer sub-categories (such as Company, Institute, and Political Party for Organization; and Country, Province, and City for Location) and adding some new types of NE categories (Position Title, Product, Event, and Natural Object).","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"We call these N \u2212 1 words state.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"Our first model is GTplain, a PCFG learned from the treebank after removing all functional features from the syntactic categories.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"Limitations There are several limitations in the methods.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"Hence, we use the bootstrap resampling method described by Koehn (2004).","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"The method just described segments dictionary words, but as noted in Section 1, there are several classes of words that should be handled that are not found in a standard dictionary.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"More recently, (Riloff and Jones 99) describe a method they term &quot;mutual bootstrapping&quot; for simultaneously constructing a lexicon and contextual extraction patterns.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Finally, as\u00c2\u00ad suming a simple bigram backoff model, we can derive the probability estimate for the particular unseen word i\u00c2\u00a51J1l.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Statistical NERs usually find the sequence of tags that maximizes the probability , where is the sequence of words in a sentence, and is the sequence of named-entity tags assigned to the words in . Attempts have been made to use global information (e.g., the same named entity occurring in different sentences of the same document), but they usually consist of incorporating an additional classifier, which tries to correct the errors in the output of a first NER (Mikheev et al., 1998; Borthwick, 1999).","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"5 60.6 Table 3: Multilingual Results: We report token-level one-to-one and many-to-one accuracy on a variety of languages under several experimental settings (Section 5).","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"IRSTLM 5.60.02 (Federico et al., 2008) is a sorted trie implementation designed for lower memory consumption.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"The ratio of buckets to entries is controlled by space multiplier m > 1.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Not all the layers have been produced for all the texts yet.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Though we are not able to calculate their memory usage on our model, results reported in their paper suggest lower memory consumption than TRIE on large-scale models, at the expense of CPU time.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"To facilitate comparison with previous work, we exhaustively evaluate this grammar and two other parsing models when gold segmentation is assumed (\u00c2\u00a75).","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"While building a machine translation system is a serious undertaking, in future we hope to attract more newcomers to the field by keeping the barrier of entry as low as possible.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"However, the characterization given in the main body of the text is correct sufficiently often to be useful.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"This range is collapsed to a number of buckets, typically by taking the hash modulo the number of buckets.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"The normalization on a per-judge basis gave very similar ranking, only slightly less consistent with the ranking from the pairwise comparisons.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"This alters generation of T as follows: n P (T |\u00cf\u0088) = n P (Ti|\u00cf\u0088) i=1 Note that this distribution captures the frequency of a tag across word types, as opposed to tokens.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Chris Dyer integrated the code into cdec.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"The question is how to normalize the probabilities in such a way that smaller groupings have a better shot at winning.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"Covering the first uncovered position in the source sentence, we use the language model probability p(ej$; $).","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"The semantic agreement KS eliminates some candidates, but also provides positive evidence in one case: if the candidate and anaphor both have semantic tags human, company, date, or location that were assigned via NER or the manually labeled dictionary entries.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Recall.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Wu and Fung introduce an evaluation method they call nk-blind.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"That is, given a choice between segmenting a sequence abc into abc and ab, c, the former will always be picked so long as its cost does not exceed the summed costs of ab and c: while; it is possible for abc to be so costly as to preclude the larger grouping, this will certainly not usually be the case.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"We feel that information from a whole corpus might turn out to be noisy if the documents in the corpus are not of the same genre.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"Only 2 link in the CC- domain (buy-purchase, acquire-acquisition) and 2 links (trader-dealer and head-chief) in the PC- domain are found in the same synset of Word- Net 2.1 (http:\/\/wordnet.princeton.edu\/).","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"07 80.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"Our experimental setup therefore is designed to serve two goals.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"6) are noisy, the results confirm that label propagation within the foreign language part of the graph adds significant quality for every language.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"They contain about 200M words (25M, 110M, 40M and 19M words, respectively).","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"Indeed there are several open issues.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"We consider the unsupervised POS induction problem without the use of a tagging dictionary.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"In the labeled version of these metrics (L) both heads and arc labels must be correct, while the unlabeled version (U) only considers heads.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"However, we do not explore this possibility in the current work.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"One class comprises words derived by productive morphologi\u00c2\u00ad cal processes, such as plural noun formation using the suffix ir, menD.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"Phrase tables were extracted from the IN and OUT training corpora (not the dev as was used for instance weighting models), and phrase pairs in the intersection of the IN and OUT phrase tables were used as positive examples, with two alternate definitions of negative examples: The classifier trained using the 2nd definition had higher accuracy on a development set.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"But diacritics are not present in unvocalized text, which is the standard form of, e.g., news media documents.3 VBD she added VP PUNC S VP VBP NP ...","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"This is a simple and effective alternative to setting weights discriminatively to maximize a metric such as BLEU.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"In our second model GTvpi we also distinguished finite and non-finite verbs and VPs as 10Lattice parsing can be performed by special initialization of the chart in a CKY parser (Chappelier et al., 1999).","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"We present several variations for the lexical component P (T , W |\u00cf\u0088), each adding more complex pa rameterizations.","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"The first experiment uses data from two dependency treebanks.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"Finally, we note that Jiang\u2019s instance-weighting framework is broader than we have presented above, encompassing among other possibilities the use of unlabelled IN data, which is applicable to SMT settings where source-only IN corpora are available.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Most of these groups follow a phrase-based statistical approach to machine translation.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"Search t0 CPU time #search mWER Method [sec] error [%] QmS 0.0 0.07 108 42:6 1.0 0.13 85 37:8 2.5 0.35 44 36:6 5.0 1.92 4 34:6 10.0 10.6 0 34:5 IbmS 0.0 0.14 108 43:4 1.0 0.3 84 39:5 2.5 0.8 45 39:1 5.0 4.99 7 38:3 10.0 28.52 0 38:2 Table 6 shows example translations obtained by the three different approaches.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"Examining the word fidanzato for the \u201cNo LP\u201d and \u201cWith LP\u201d models is particularly instructive.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"Precision is the portion of hypothesized constituents that are correct and recall is the portion of the Treebank constituents that are hypothesized.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"However, there are several reasons why this approach will not in general work: 1.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Systems that generally do better than others will receive a positive average normalizedjudgement per sentence.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"t(i).","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"(2009), who also incorporate a sparsity constraint, but does via altering the model objective using posterior regularization.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"Since each composition operation is linear and nonerasing, a bounded sequences of substrings associated with the resulting structure is obtained by combining the substrings in each of its arguments using only the concatenation operation, including each substring exactly once.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"(S2) The burglar fired the gun three times and fled.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"The approach gains leverage from natural redundancy in the data: for many named-entity instances both the spelling of the name and the context in which it appears are sufficient to determine its type.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Now, for this application one might be tempted to simply bypass the segmentation problem and pronounce the text character-by-character.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"This leads to the best reported performance for robust non-projective parsing of Czech.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"For developing these mechanisms, the possibility to feed in hand-annotated information is very useful.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"(2010) and the posterior regular- ization HMM of Grac\u00c2\u00b8a et al.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"The larger sets are more accurate than the small sets.","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"The 3rd block contains the mixture baselines.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"Segmentation errors cascade into the parsing phase, placing an artificial limit on parsing performance.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"TPT Germann et al. (2009) describe tries with better locality properties, but did not release code.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":":zhong1 : 0.0 tjl :huo2 :0.0 (R:spub:\/ic of Ch:ina) + .,_,...I : jlong4 :0.0 (mUifaty genG181) 0 \u00c2\u00a3: _NC: 40.0 Figure 3 Partial Chinese Lexicon (NC = noun; NP = proper noun).c=- - I \u00e2\u0080\u00a2=- :il: .;ss:;zhangt \u00e2\u0080\u00a2 '-:.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"In this particular case, all English vertices are labeled as nouns by the supervised tagger.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"We smooth Prf(p \u2014* (s, p)) for rare and OOV segments (s E l, l E L, s unseen) using a \u201cper-tag\u201d probability distribution over rare segments which we estimate using relative frequency estimates for once-occurring segments.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"The weak hypothesis chosen was then restricted to be a predictor in favor of this label.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"(Brandt 1996) extended these ideas toward a conception of kommunikative Gewichtung (\u00e2\u0080\u0098communicative-weight assignment\u00e2\u0080\u0099).","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"To investigate the influence of these factors, we analyze Modern Standard Arabic (henceforth MSA, or simply \u00e2\u0080\u009cArabic\u00e2\u0080\u009d) because of the unusual opportunity it presents for comparison to English parsing results.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Each visited entry wni stores backoff b(wni ).","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"Given around 90,000 unlabeled examples, the methods described in this paper classify names with over 91% accuracy.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Training examples are generated automatically by identifying noun phrases that can be easily resolved with their antecedents using lexical and syntactic heuristics.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"Evaluation Metrics We report three metrics to evaluate tagging performance.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"As the two NE categories are the same, we can\u00e2\u0080\u0099t differentiate phrases with different orders of par ticipants \u00e2\u0080\u0093 whether the buying company or the to-be-bought company comes first.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"(2009) on Portuguese (Grac\u00c2\u00b8a et al.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"We will return to these issues in the discussion section.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"In addition to a heuristic based on decision list learning, we also presented a boosting-like framework that builds on ideas from (Blum and Mitchell 98).","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"We also mark all nodes that dominate an SVO configuration (containsSVO).","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"A TAG consists of a finite set of elementary trees that are either initial trees or auxiliary trees.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"Cohen and Smith (2007) later on based a system for joint inference on factored, independent, morphological and syntactic components of which scores are combined to cater for the joint inference task.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"We divide up each test set into blocks of 20 sentences (100 blocks for the in-domain test set, 53 blocks for the out-of-domain test set), check for each block, if one system has a higher BLEU score than the other, and then use the sign test.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"There is twice as much language modelling data, since training data for the machine translation system is filtered against sentences of length larger than 40 words.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Unigram lookup is dense so we use an array of probability and backoff values.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"Across all languages, high performance can be attained by selecting a single tag per word type.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"The NIST06 and NIST08 evaluation sets were used for testing.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"Then, the German infinitive 'besuchen' and the negation particle 'nicht' are translated.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"Human evaluation is one way to distinguish between the two cases.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"3.1 Word ReOrdering with Verbgroup.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":".","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"For example, given a sequence F1G1G2, where F1 is a legal single-hanzi family name, and Plural Nouns X g 0 g \"' X X 0 T!i c\"'.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"For the translation experiments, Eq. 2 is recursively evaluated.","label":0},{"source_text":"Here we present two algorithms.","target_text":"This PP modifies another NP, whose head is a singular noun.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Finally, we add \u00e2\u0080\u009cDT\u00e2\u0080\u009d to the tags for definite nouns and adjectives (Kulick et al., 2006).","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"\u00e2\u0080\u00a2 Some tools would allow for the desired annotation mode, but are so complicated (they can be used for many other purposes as well) that annotators take a long time getting used to them.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"Our second contribution is to apply instance weighting at the level of phrase pairs.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"Tag set As is standard, for all experiments, we set the number of latent model tag states to the size of the annotated tag set.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"7 Unlike Dickinson (2005), we strip traces and only con-.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"Let say, if we find one system doing better on 20 of the blocks, and worse on 80 of the blocks, is it significantly worse?","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"For brevity we omit the segments from the analysis, and so analysis of the form \u201cfmnh\u201d as f\/REL mnh\/VB is represented simply as REL VB.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"47 78.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"The gender, number, and scoping KSs eliminate candidates from consideration.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"There are simply not enough votes remaining to allow any of the crossing structures to enter the hypothesized constituent set.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Caching for IRSTLM is smaller at 0.09 GB resident memory, though it supports only a single thread.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"Obviously \u00e2\u0080\u009cLotus\u00e2\u0080\u009d is part of the following clause rather than being the object of \u00e2\u0080\u009cestimates\u00e2\u0080\u009d and the extracted instance makes no sense.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"The simplest approach involves scoring the various analyses by costs based on word frequency, and picking the lowest cost path; variants of this approach have been described in Chang, Chen, and Chen (1991) and Chang and Chen (1993).","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"The set of candidate constituents comes from the union of all the constituents suggested by the member parsers.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Links can be of two different kinds: anaphoric or bridging (definite noun phrases picking up an antecedent via world-knowledge).","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"Here we push the single-framework conjecture across the board and present a single model that performs morphological segmentation and syntactic disambiguation in a fully generative framework.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"The second term is a regularizer and encourages all type marginals to be uniform to the extent that is allowed by the first two terms (cf. maximum entropy principle).","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"4.3 Morphological Analysis.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"12 One class of full personal names that this characterization does not cover are married women's names.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"As a first step towards modeling transliterated names, we have collected all hanzi occurring more than once in the roughly 750 foreign names in our dictionary, and we estimate the probabil\u00c2\u00ad ity of occurrence of each hanzi in a transliteration (pTN(hanzi;)) using the maximum likelihood estimate.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"Our full model outperforms the \u201cNo LP\u201d setting because it has better vocabulary coverage and allows the extraction of a larger set of constraint features.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Our System Wang, Li, and Chang a. 1\\!f!IP Eflltii \/1\\!f!J:P $1til I b. agm: I a m: c. 5 Bf is Bf 1 d. \"*:t: w _t ff 1 \"* :t: w_tff 1 g., , Transliteration\/Translation chen2zhongl-shenl qu3 'music by Chen Zhongshen ' huang2rong2 youlyoul de dao4 'Huang Rong said soberly' zhangl qun2 Zhang Qun xian4zhang3 you2qingl shang4ren2 hou4 'after the county president You Qing had assumed the position' lin2 quan2 'Lin Quan' wang2jian4 'Wang Jian' oulyang2-ke4 'Ouyang Ke' yinl qi2 bu4 ke2neng2 rong2xu3 tai2du2 er2 'because it cannot permit Taiwan Independence so' silfa3-yuan4zhang3 lin2yang2-gang3 'president of the Judicial Yuan, Lin Yanggang' lin2zhangl-hu2 jiangl zuo4 xian4chang3 jie3shuol 'Lin Zhanghu will give an ex\u00c2\u00ad planation live' jin4\/iang3 nian2 nei4 sa3 xia4 de jinlqian2 hui4 ting2zhi3 'in two years the distributed money will stop' gaoltangl da4chi2 ye1zi0 fen3 'chicken stock, a tablespoon of coconut flakes' you2qingl ru4zhu3 xian4fu3 lwu4 'after You Qing headed the county government' Table 5 Performance on morphological analysis.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"3.1 Word ReOrdering with Verbgroup.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"We first note that the accuracy results of our system are overall higher on their setup, on all measures, indicating that theirs may be an easier dataset.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"1 61.7 37.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"Our analysis identifies three key factors driving our performance gain: 1) selecting a model structure which directly encodes tag sparsity, 2) a type-level prior on tag assignments, and 3) a straightforward na\u00c2\u00a8\u00c4\u00b1veBayes approach to incorporate features.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"This result suggests that all of contextual role KSs can provide useful information for resolving anaphora.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"The numbers falling into the location, person, organization categories were 186, 289 and 402 respectively.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"However, 1 http:\/\/maxent.sourceforge.net 3.2 Testing.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"We start with noun features since written Arabic contains a very high proportion of NPs.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"The predominate focus of building systems that translate into English has ignored so far the difficult issues of generating rich morphology which may not be determined solely by local context.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Many hanzi have more than one pronunciation, where the correct.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"JA DE G O L D G R AS S SI C K NE SS DE AT H R A T 14.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"\u00e2\u0080\u009cAgree\u00e2\u0080\u009d is a subject control verb, which dominates another verb whose subject is the same as that of \u00e2\u0080\u009cagree\u00e2\u0080\u009d; the latter verb is generally the one of interest for extraction.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"After tagging a large corpus with an automatic NE tagger, the method tries to find sets of paraphrases automatically without being given a seed phrase or any kinds of cue.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"For a given partial hypothesis (C; j), the order in which the cities in C have been visited can be ignored (except j), only the score for the best path reaching j has to be stored.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"The second concerns the methods used (if any) to ex\u00c2\u00ad tend the lexicon beyond the static list of entries provided by the machine-readable dictionary upon which it is based.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","label":1},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"SRILM\u2019s compact variant, IRSTLM, MITLM, and BerkeleyLM\u2019s sorted variant are all based on this technique.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"Participants and other volunteers contributed about 180 hours of labor in the manual evaluation.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"We performed three experiments to evaluate our techniques.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"Finally we show the combining techniques degrade very little when a poor parser is added to the set.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"For all other recursive NPs, we add a common annotation to the POS tag of the head (recursiveNPHead).","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"If two nouns have mutually exclusive semantic classes, then they cannot be coreferent.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"Dictionaries: Due to the limited amount of training material, name dictionaries have been found to be useful in the named entity task.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"(f1; ;mg n fl1; l2g ; l) 4 (f1; ;m \u00f4\u0080\u0080\u0080 1g n fl1; l2; l3g ; l0) !","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"One knowledge source, called WordSemCFSem, is analogous to CFLex: it checks whether the anaphor and candidate antecedent are substitutable for one another, but based on their semantic classes instead of the words themselves.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"Our results suggest that it is possible to learn accurate POS taggers for languages which do not have any annotated data, but have translations into a resource-rich language.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"Surprisingly, the non-parametric switching technique also exhibited robust behaviour in this situation.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"We train and test on the CoNLL-X training set.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"However, 1 http:\/\/maxent.sourceforge.net 3.2 Testing.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"In this specific case, as these two titles could fill the same column of an IE table, we regarded them as paraphrases for the evaluation.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"a classifier.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"All systems (except for Systran, which was not tuned to Europarl) did considerably worse on outof-domain training data.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"This work was supported in part under the GALE program of the Defense Advanced Research Projects Agency, Contract No.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"Upon identifying an anaphoric expression (currently restricted to: pronouns, prepositional adverbs, definite noun phrases), the an- notator first marks the antecedent expression (currently restricted to: various kinds of noun phrases, prepositional phrases, verb phrases, sentences) and then establishes the link between the two.","label":0},{"source_text":"There is no global pruning.","target_text":"Additionally, it works about 3 times as fast as the IBM style search.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"The results in Table 2 were achieved on the development set.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Our code is thread-safe, and integrated into the Moses, cdec, and Joshua translation systems.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"The bias of automatic methods in favor of statistical systems seems to be less pronounced on out-of-domain test data.","label":1},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Only 2 link in the CC- domain (buy-purchase, acquire-acquisition) and 2 links (trader-dealer and head-chief) in the PC- domain are found in the same synset of Word- Net 2.1 (http:\/\/wordnet.princeton.edu\/).","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"input: source string f1:::fj :::fJ initialization for each cardinality c = 1; 2; ; J do for each pair (C; j), where j 2 C and jCj = c do for each target word e 2 E Qe0 (e; C; j) = p(fj je) max \u00c3\u0086;e00 j02Cnfjg fp(jjj0; J) p(\u00c3\u0086) p\u00c3\u0086(eje0; e00) Qe00 (e0;C n fjg; j0)g words fj in the input string of length J. For the final translation each source position is considered exactly once.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"In many cases, inspection of either the spelling or context alone is sufficient to classify an example.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Nodes in the trie are based on arrays sorted by vocabulary identifier.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"We have shown that the maximum entropy framework is able to use global information directly.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"SSER: subjective sentence error rate: For a more detailed analysis, the translations are judged by a human test person.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"We claim that no particular morphological segmentation is a-priory more likely for surface forms before exploring the compositional nature of syntactic structures, including manifestations of various long-distance dependencies.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"Employing a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions is not only theoretically clean and linguistically justified and but also probabilistically apropriate and empirically sound.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"If a phrase does not contain any keywords, the phrase is discarded.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"The phrases have to be the expressions of length less than 5 chunks, appear between two NEs.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"Experiments are presented in section 4.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"On the one hand, the type-level error rate is not calibrated for the number of n-grams in the sample.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"All experiments use ATB parts 1\u00e2\u0080\u00933 divided according to the canonical split suggested by Chiang et al.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"6) are noisy, the results confirm that label propagation within the foreign language part of the graph adds significant quality for every language.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Upon identifying an anaphoric expression (currently restricted to: pronouns, prepositional adverbs, definite noun phrases), the an- notator first marks the antecedent expression (currently restricted to: various kinds of noun phrases, prepositional phrases, verb phrases, sentences) and then establishes the link between the two.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"AdaBoost finds a weighted combination of simple (weak) classifiers, where the weights are chosen to minimize a function that bounds the classification error on a set of training examples.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"If no candidate satisfies this condition (which is often the case), then the anaphor is left unresolved.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"The difference in precision between similarity and Bayes switching techniques is significant, but the difference in recall is not.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"We then show that the combining techniques presented above give better parsing accuracy than any of the individual parsers.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"information structure.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"3 58.3 40.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Again, we can compute average scores for all systems for the different language pairs (Figure 6).","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"The first experiment uses data from two dependency treebanks.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Fortunately, there are only a few hundred hanzi that are particularly common in transliterations; indeed, the commonest ones, such as E. bal, m er3, and iij al are often clear indicators that a sequence of hanzi containing them is foreign: even a name like !:i*m xia4mi3-er3 'Shamir,' which is a legal Chi\u00c2\u00ad nese personal name, retains a foreign flavor because of liM.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"It is closer to the smaller value of precision and recall when there is a large skew in their values.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Unigrams also have 64-bit overhead for vocabulary lookup.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"For robustness reasons, the parser may output a set of dependency trees instead of a single tree. most dependent of the next input token, dependency type features are limited to tokens on the stack.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"Both of the switching techniques, as well as the parametric hybridization technique were also shown to be robust when a poor parser was introduced into the experiments.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"In our model there are no such hyper-parameters, and the performance is the result of truly joint disambiguation. sults.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"And if one is interested in TIS, one would probably consider the single orthographic word ACL to consist of three phonological words-lei s'i d\/-corresponding to the pronunciation of each of the letters in the acronym.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"From this we see that a finer-grained model for parser combination, at least for the features we have examined, will not give us any additional power.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"Performance typically stabilizes across languages after only a few number of iterations.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"13.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"This departure from the traditional token-based tagging approach allows us to explicitly capture type- level distributional properties of valid POS tag as signments as part of the model.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"Each set is assigned two values: belief and plausibility.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"(Berland and Charniak 99) describe a method for extracting parts of objects from wholes (e.g., &quot;speedometer&quot; from &quot;car&quot;) from a large corpus using hand-crafted patterns.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"The likelihood of the observed data under the model is where P(yi, xi) is defined as in (9).","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"First, we aim to explicitly characterize examples from OUT as belonging to general language or not.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"May 1995).","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Thus in a two-hanzi word like lflli?J zhong1guo2 (middle country) 'China' there are two syllables, and at the same time two morphemes.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"The normalization factor plays an important role in the AdaBoost algorithm.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"2.1 Reliable Case Resolutions.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"Special thanks to Jan Haji\u02c7c and Matthias Trautner Kromann for assistance with the Czech and Danish data, respectively, and to Jan Haji\u02c7c, Tom\u00b4a\u02c7s Holan, Dan Zeman and three anonymous reviewers for valuable comments on a preliminary version of the paper.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"This model is easily incorporated into the segmenter by building a WFST restrict\u00c2\u00ad ing the names to the four licit types, with costs on the arcs for any particular name summing to an estimate of the cost of that name.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"3.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"We consider two variants of Berg-Kirkpatrick et al.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"The second setting uses the news-related subcorpora for the NIST09 MT Chinese to English evaluation8 as IN, and the remaining NIST parallel Chinese\/English corpora (UN, Hong Kong Laws, and Hong Kong Hansard) as OUT.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"The paper is organized as follows: Section 2 explains the different layers of annotation that have been produced or are being produced.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"The algorithm works due to the fact that not all permutations of cities have to be considered explicitly.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"2.5 Connectives with scopes.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Other good classes include JADE and GOLD; other bad classes are DEATH and RAT.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"The graph was constructed using 2 million trigrams; we chose these by truncating the parallel datasets up to the number of sentence pairs that contained 2 million trigrams.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"For instance, the sentence Similar improvements in haemoglobin levels were reported in the scientific literature for other epoetins would likely be considered domain-specific despite the presence of general phrases like were reported in.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"Put another way, the minimum of Equ.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"more frequently than is done in English.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"83 77.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"We use N(u) to denote the neighborhood of vertex u, and fixed n = 5 in our experiments.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"In this example there are four \"input characters,\" A, B, C and D, and these map respectively to four \"pronunciations\" a, b, c and d. Furthermore, there are four \"words\" represented in the dictionary.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"This enables us to build a high performance NER without using separate classifiers to take care of global consistency or complex formulation on smoothing and backoff models (Bikel et al., 1997).","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"For effectively annotating connectives\/scopes, we found that existing annotation tools were not well-suited, for two reasons: \u00e2\u0080\u00a2 Some tools are dedicated to modes of annotation (e.g., tiers), which could only quite un-intuitively be used for connectives and scopes.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"We carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"BABAR applies a DempsterShafer probabilistic model to make resolutions based on evidence from the contextual role knowledge sources as well as general knowledge sources.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"The probability distribution that satisfies the above property is the one with the highest entropy.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"The only supervision is in the form of 7 seed rules (namely, that New York, California and U.S. are locations; that any name containing Mr is a person; that any name containing Incorporated is an organization; and that I.B.M. and Microsoft are organizations).","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"The tree denoting this derivation of 7 is rooted with a node labeled 7' having k subtrees for the derivations of 71, ... ,7a.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"summit Sharm (a) Al-Sheikh summit Sharm (b) DTNNP Al-Sheikh in a corpus position without a bracketing label, then we also add \u00e2\u0088\u0097n, NIL) to M. We call the set of unique n-grams with multiple labels in M the variation nuclei of C. Bracketing variation can result from either annotation errors or linguistic ambiguity.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"In order to handle the necessary word reordering as an optimization problem within our dynamic programming approach, we describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming (Held, Karp, 1962).","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"It was our hope that this competition, which included the manual and automatic evaluation of statistical systems and one rulebased commercial system, will give further insight into the relation between automatic and manual evaluation.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":".., Wn ) (obs) P T : Tag assigns (T1 ,.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"None of the parsers produce parses with crossing brackets, so none of them votes for both of the assumed constituents.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"(Hartmann 1984), for example, used the term Reliefgebung to characterize the distibution of main and minor information in texts (similar to the notion of nuclearity in RST).","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"The accuracy is calculated as the ratio of the number of paraphrases to the total number of phrases in the set.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"On MUC6 data, MENERGI also achieves performance comparable to IdentiFinder when trained on similar amount of training data.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Given the closeness of most systems and the wide over-lapping confidence intervals it is hard to make strong statements about the correlation between human judgements and automatic scoring methods such as BLEU.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"While there might be some controversy about the exact definition of such a tagset, these 12 categories cover the most frequent part-of-speech and exist in one form or another in all of the languages that we studied.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"This is true of the widely used link grammar parser for English (Sleator and Temperley, 1993), which uses a dependency grammar of sorts, the probabilistic dependency parser of Eisner (1996), and more recently proposed deterministic dependency parsers (Yamada and Matsumoto, 2003; Nivre et al., 2004).","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"The details are given in (Och and Ney, 2000).","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"For the purposes of EM, the &quot;observed&quot; data is {(xi, Ya\u2022 \u2022 \u2022 (xrn, Yrn), xfil, and the hidden data is {ym+i y}.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"First, we will describe their method and compare it with our method.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Evaluation of the Segmentation as a Whole.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Since we could not bias the subjects towards a particular segmentation and did not presume linguistic sophistication on their part, the instructions were simple: subjects were to mark all places they might plausibly pause if they were reading the text aloud.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Obviously, the presence of a title after a potential name N increases the probability that N is in fact a name.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"Sometimes extracted phrases by themselves are not meaningful to consider without context, but we set the following criteria.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"The web-based Annis imports data in a variety of XML formats and tagsets and displays it in a tier-orientedway (optionally, trees can be drawn more ele gantly in a separate window).","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"This drastic tree manipulation is not appropriate for situations in which we want to assign particular structures to sentences.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Given a set of n sentences, we can compute the sample mean x\ufffd and sample variance s2 of the individual sentence judgements xi: The extend of the confidence interval [x\u2212d, x+df can be computed by d = 1.96 \u00b7\ufffdn (6) Pairwise Comparison: As for the automatic evaluation metric, we want to be able to rank different systems against each other, for which we need assessments of statistical significance on the differences between a pair of systems.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"Simply assigning to each word its most frequent associated tag in a corpus achieves 94.6% accuracy on the WSJ portion of the Penn Treebank.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"This measure has the advantage of being completely automatic.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"For RandLM, we used the settings in the documentation: 8 bits per value and false positive probability 1 256.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"Our second point of comparison is with Grac\u00c2\u00b8a et al.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"The human judges were presented with the following definition of adequacy and fluency, but no additional instructions:","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"In this subsection, we will report the results of the experiment, in terms of the number of words, phrases or clusters.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"Corporate Suffixes and Person Prefixes of Other Occurrences (CSPP): If McCann has been seen as Mr. McCann somewhere else in the document, then one would like to give person a higher probability than organization.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"Across eight European languages, our approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"First, we use a novel graph-based framework for projecting syntactic information across language boundaries.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Saving state allows our code to walk the data structure exactly once per query.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"To quantize, we use the binning method (Federico and Bertoldi, 2006) that sorts values, divides into equally sized bins, and averages within each bin.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"The sources of our dictionaries are listed in Table 2.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"Using a wide-coverage morphological analyzer based on (Itai et al., 2006) should cater for a better coverage, and incorporating lexical probabilities learned from a big (unannotated) corpus (cf.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"The type of alignment we have considered so far requires the same length for source and target sentence, i.e. I = J. Evidently, this is an unrealistic assumption, therefore we extend the concept of inverted alignments as follows: When adding a new position to the coverage set C, we might generate either \u00c3\u0086 = 0 or \u00c3\u0086 = 1 new target words.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"Thus we opted not to take the step of creating more precise written annotation guidelines (as (Carlson, Marcu 2001) did for English), which would then allow for measuring inter-annotator agreement.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"On several languages, we report performance exceeding that of state-of-the art systems.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"However, for multinomial models like our LMs and TMs, there is a one to one correspondence between instances and features, eg the correspondence between a phrase pair (s, t) and its conditional multinomial probability p(s1t).","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"Acknowledgments We thank Meni Adler and Michael Elhadad (BGU) for helpful comments and discussion.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"During coreference resolution, the caseframe network provides evidence that an anaphor and prior noun phrase might be coreferent.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Thus we have some confidence that our own performance is at least as good as that of Chang et al.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"The annotator can then \u00e2\u0080\u009cclick away\u00e2\u0080\u009d those words that are here not used as connectives (such as the conjunction und (\u00e2\u0080\u0098and\u00e2\u0080\u0099) used in lists, or many adverbials that are ambiguous between connective and discourse particle).","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"At present, the \u00e2\u0080\u0098Potsdam Commentary Corpus\u00e2\u0080\u0099 (henceforth \u00e2\u0080\u0098PCC\u00e2\u0080\u0099 for short) consists of 170 commentaries from Ma\u00c2\u00a8rkische Allgemeine Zeitung, a German regional daily.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"For all languages, the vocabulary sizes increase by several thousand words.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"The high 1 tone of J1l would not normally neutralize in this fashion if it were functioning as a word on its own.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"This class-based model gives reasonable results: for six radical classes, Table 1 gives the estimated cost for an unseen hanzi in the class occurring as the second hanzi in a double GIVEN name.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"If the same pair of NE instances is used with different phrases, these phrases are likely to be paraphrases.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"We will evaluate various specific aspects of the segmentation, as well as the overall segmentation per\u00c2\u00ad formance.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"Thus, TAG's can not give analyses in which dependencies between arbitrarily large branches exist.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"Employing a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions is not only theoretically clean and linguistically justified and but also probabilistically apropriate and empirically sound.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Quite often, though, these directives fulfill the goal of increasing annotator agreement without in fact settling the theoretical question; i.e., the directives are clear but not always very well motivated.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"For example, in Northern dialects (such as Beijing), a full tone (1, 2, 3, or 4) is changed to a neutral tone (0) in the final syllable of many words: Jll donglgual 'winter melon' is often pronounced donglguaO.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"All our results are obtained by using only the official training data provided by the MUC conferences.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"The supervised POS tagging accuracies (on this tagset) are shown in the last row of Table 2.","label":0},{"source_text":"Here we present two algorithms.","target_text":"The algorithm, called CoBoost, has the advantage of being more general than the decision-list learning alInput: (xi , yi), , (xim, ) ; x, E 2x, yi = +1 Initialize Di (i) = 1\/m.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Notice that the CC-domain is a special case.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Clearly it is possible to write a rule that states that if an analysis Modal+ Verb is available, then that is to be preferred over Noun+ Verb: such a rule could be stated in terms of (finite-state) local grammars in the sense of Mohri (1993).","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Mohri [1995]) shows promise for improving this situation.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"First, from a large corpus, we extract all the NE instance pairs.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"This is the only important case, because otherwise the simple majority combining technique would pick the correct constituent.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"The segmentation chosen is the best path through the WFST, shown in (d).","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"The text type are editorials instead of speech transcripts.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"In principle a feature could be an arbitrary predicate of the (spelling, context) pair; for reasons that will become clear, features are limited to querying either the spelling or context alone.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"input: source string f1:::fj :::fJ initialization for each cardinality c = 1; 2; ; J do for each pair (C; j), where j 2 C and jCj = c do for each target word e 2 E Qe0 (e; C; j) = p(fj je) max \u00c3\u0086;e00 j02Cnfjg fp(jjj0; J) p(\u00c3\u0086) p\u00c3\u0086(eje0; e00) Qe00 (e0;C n fjg; j0)g words fj in the input string of length J. For the final translation each source position is considered exactly once.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"rently, some annotations (in particular the connectives and scopes) have already moved beyond the core corpus; the others will grow step by step.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"4 65.9 48.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"Named Entity Recognition: A Maximum Entropy Approach Using Global Information","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"The input for the segmentation task is however highly ambiguous for Semitic languages, and surface forms (tokens) may admit multiple possible analyses as in (BarHaim et al., 2007; Adler and Elhadad, 2006).","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"However, the next step is clearly different.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"They are set to fixed constants.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"The observed performance gains, coupled with the simplicity of model implementation, makes it a compelling alternative to existing more complex counterparts.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"The 2(Yarowsky 95) describes the use of more sophisticated smoothing methods.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"Also, we don\u00e2\u0080\u0099t know how many such paraphrase sets are necessary to cover even some everyday things or events.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"This group consists of (3 total number of possible zones) features.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"The less favored reading may be selected in certain contexts, however; in the case of , for example, the nominal reading jiang4 will be selected if there is morphological information, such as a following plural affix ir, menD that renders the nominal reading likely, as we shall see in Section 4.3.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"We show translation results for three approaches: the monotone search (MonS), where no word reordering is allowed (Tillmann, 1997), the quasimonotone search (QmS) as presented in this paper and the IBM style (IbmS) search as described in Section 3.2.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"\u2014 I would also like to point out to commissioner Liikanen that it is not easy to take a matter to a national court.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"We are given a source string fJ 1 = f1:::fj :::fJ of length J, which is to be translated into a target string eI 1 = e1:::ei:::eI of length I. Among all possible target strings, we will choose the string with the highest probability: ^eI 1 = arg max eI 1 fPr(eI 1jfJ 1 )g = arg max eI 1 fPr(eI 1) Pr(fJ 1 jeI 1)g : (1) The argmax operation denotes the search problem, i.e. the generation of the output sentence in the target language.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"The manual evaluation of scoring translation on a graded scale from 1\u20135 seems to be very hard to perform.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"It is closer to the smaller value of precision and recall when there is a large skew in their values.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Therefore, for n-gram wn1 , all leftward extensions wn0 are an adjacent block in the n + 1-gram array.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"In (1) the sequencema3lu4 cannot be resolved locally, but depends instead upon broader context; similarly in (2), the sequence :::tcai2neng2 cannot be resolved locally: 1.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"In this example there are four \"input characters,\" A, B, C and D, and these map respectively to four \"pronunciations\" a, b, c and d. Furthermore, there are four \"words\" represented in the dictionary.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"By aggregating the POS labels of the English tokens to types, we can generate label distributions for the English vertices.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"On the MUC6 data, Bikel et al.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"The method reported in this paper makes use solely of unigram probabilities, and is therefore a zeroeth-order model: the cost of a particular segmentation is estimated as the sum of the costs of the individual words in the segmentation.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"3.1 Lexicon Component.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"This is can not be the only explanation, since the discrepancy still holds, for instance, for out-of-domain French-English, where Systran receives among the best adequacy and fluency scores, but a worse BLEU score than all but one statistical system.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"The bias of automatic methods in favor of statistical systems seems to be less pronounced on out-of-domain test data.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Now, for this application one might be tempted to simply bypass the segmentation problem and pronounce the text character-by-character.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"Using the terminology of Kahane et al. (1998), we say that jedna is the syntactic head of Z, while je is its linear head in the projectivized representation.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"A typical situation is shown in Figure 1.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"We maintain a separate array for each length n containing all n-gram entries sorted in suffix order.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Since the inclusion of out-ofdomain test data was a very late decision, the participants were not informed of this.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"For derived words that occur in our corpus we can estimate these costs as we would the costs for an underived dictionary entry.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"The effect of UNIQ is very small on both data sets.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"Our results outperform strong unsupervised baselines as well as approaches that rely on direct projections, and bridge the gap between purely supervised and unsupervised POS tagging models.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"The evidence may not say anything about whether A is more likely than B, only that C and D are not likely.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"It is well know that language pairs such as EnglishGerman pose more challenges to machine translation systems than language pairs such as FrenchEnglish.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"Yet we note that the better grammars without pruning outperform the poorer grammars using this technique, indicating that the syntactic context aids, to some extent, the disambiguation of unknown tokens.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Evaluation within a set The evaluation of paraphrases within a set of phrases which share a keyword is illustrated in Figure 4.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Gan's solution depends upon a fairly sophisticated language model that attempts to find valid syntactic, semantic, and lexical relations between objects of various linguistic types (hanzi, words, phrases).","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"We evaluate our approach on eight European languages (\u00a76), and show that both our contributions provide consistent and statistically significant improvements.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"This remains as future work.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"This work was supported in part under the GALE program of the Defense Advanced Research Projects Agency, Contract No.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"Same for . In the case where the next token is a hyphen, then is also used as a feature: (init- Caps, ) is set to 1.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"TIS systems in general need to do more than simply compute the.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"For verbs we add two features.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"Ex: The government said it ...","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"For the \u00e2\u0080\u0098core\u00e2\u0080\u0099 portion of PCC, we found that on average, 35% of the coherence relations in our RST annotations are explicitly signalled by a lexical connective.6 When adding the fact that connectives are often ambiguous, one has to conclude that prospects for an automatic analysis of rhetorical structure using shallow methods (i.e., relying largely on connectives) are not bright \u00e2\u0080\u0094 but see Sections 3.2 and 3.3 below.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"The inverted alignment probability p(bijbi\u00f4\u0080\u0080\u00801; I; J) and the lexicon probability p(fbi jei) are obtained by relative frequency estimates from the Viterbi alignment path after the final training iteration.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"Although matching is done at the sentence level, this information is subsequently discarded when all matches are pooled.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"While size of the resulting transducers may seem daunting-the segmenter described here, as it is used in the Bell Labs Mandarin TTS system has about 32,000 states and 209,000 arcs-recent work on minimization of weighted machines and transducers (cf.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"The segmenter will give both analyses :1 cai2 neng2 'just be able,' and ?]cai2neng2 'talent,' but the latter analysis is preferred since splitting these two morphemes is generally more costly than grouping them.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"Since the inclusion of out-ofdomain test data was a very late decision, the participants were not informed of this.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"For humans, this characteristic can impede the acquisition of literacy.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"Apart from MERT difficulties, a conceptual problem with log-linear combination is that it multiplies feature probabilities, essentially forcing different features to agree on high-scoring candidates.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"8 66.4 52.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"Eight out of the thirteen errors in the high frequency phrases in the CC-domain are the phrases in \u00e2\u0080\u009cagree\u00e2\u0080\u009d.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"In this section, we extend state to optimize left-to-right queries.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"One may argue with these efforts on normalization, and ultimately their value should be assessed by assessing their impact on inter-annotator agreement.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"The two new terms force the two classifiers to agree, as much as possible, on the unlabeled examples.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"52 15.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"(Bar-Haim et al., 2007; Habash and Rambow, 2005)) and probabilities are assigned to different analyses in accordance with the likelihood of their tags (e.g., \u201cfmnh is 30% likely to be tagged NN and 70% likely to be tagged REL+VB\u201d).","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Finally, Section 5 explains how BABAR relates to previous work, and Section 6 summarizes our conclusions.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"This paper discusses the use of unlabeled examples for the problem of named entity classification.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"97 81.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"All notions of word, with the exception of the orthographic word, are as relevant in Chinese as they are in English, and just as is the case in other languages, a word in Chinese may correspond to one or more symbols in the orthog 1 For a related approach to the problem of word-segrnention in Japanese, see Nagata (1994), inter alia..","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"AdaBoost finds a weighted combination of simple (weak) classifiers, where the weights are chosen to minimize a function that bounds the classification error on a set of training examples.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Many hanzi have more than one pronunciation, where the correct.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"(student+plural) 'students,' which is derived by the affixation of the plural affix f, menD to the nounxue2shengl.","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"We incorporate instance weighting into a mixture-model framework, and find that it yields consistent improvements over a wide range of baselines.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"The implementation is In a similar effort, (G\u00c2\u00a8otze 2003) developed a proposal for the theory-neutral annotation of information structure (IS) \u00e2\u0080\u0094 a notoriously difficult area with plenty of conflicting and overlapping terminological conceptions.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"Our work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"An important subproblem of language model storage is therefore sparse mapping: storing values for sparse keys using little memory then retrieving values given keys using little time.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"For a word whose initCaps might be due to its position rather than its meaning (in headlines, first word of a sentence, etc), the case information of other occurrences might be more accurate than its own.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"A final alternate approach would be to combine weighted joint frequencies rather than conditional estimates, ie: cI(s, t) + w,\\(s, t)co(, s, t), suitably normalized.5 Such an approach could be simulated by a MAP-style combination in which separate 0(t) values were maintained for each t. This would make the model more powerful, but at the cost of having to learn to downweight OUT separately for each t, which we suspect would require more training data for reliable performance.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"Links can be of two different kinds: anaphoric or bridging (definite noun phrases picking up an antecedent via world-knowledge).","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"The negative logarithm of t0 is reported.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"Again, we can compute average scores for all systems for the different language pairs (Figure 6).","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"9 50.2 +P RI OR be st me dia n 47.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Using cn to denote the number of n-grams, total memory consumption of TRIE, in bits, is plus quantization tables, if used.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"2.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"25 16.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"Dagan and Itai (Dagan and Itai, 1990) experimented with co-occurrence statistics that are similar to our lexical caseframe expectations.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"A graphical depiction of our model as well as a summary of random variables and parameters can be found in Figure 1.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"There is a guarantee of no crossing brackets but there is no guarantee that a constituent in the tree has the same children as it had in any of the three original parses.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"The learning task is to find two classifiers : 2x1 { \u20141, +1} 12 : 2x2 { \u20141, +1} such that (x1,) = f2(x2,t) = yt for examples i = 1, , m, and f1 (x1,) = f2 (x2,t) as often as possible on examples i = m + 1, ,n. To achieve this goal we extend the auxiliary function that bounds the training error (see Equ.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"But we follow the more direct adaptation of Evalb suggested by Tsarfaty (2006), who viewed exact segmentation as the ultimate goal.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":".., Tn ) T W \u00cf\u0086 E w : Token word seqs (obs) t : Token tag assigns (det by T ) PARAMETERS \u00cf\u0088 : Lexicon parameters \u00ce\u00b8 : Token word emission parameters \u00cf\u0086 : Token tag transition parameters \u00cf\u0086 \u00cf\u0086 t1 t2 \u00ce\u00b8 \u00ce\u00b8 w1 w2 K \u00cf\u0086 T tm O K \u00ce\u00b8 E wN m N N Figure 1: Graphical depiction of our model and summary of latent variables and parameters.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"To optimize left-to-right queries, we extend state to store backoff information: where m is the minimal context from Section 4.1 and b is the backoff penalty.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"For more on the participating systems, please refer to the respective system description in the proceedings of the workshop.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"These words are in turn highly ambiguous, breaking the assumption underlying most parsers that the yield of a tree for a given sentence is known in advance.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"In practice, this sparsity constraint is difficult to incorporate in a traditional POS induction system (Me\u00c2\u00b4rialdo, 1994; Johnson, 2007; Gao and Johnson, 2008; Grac\u00c2\u00b8a et al., 2009; Berg-Kirkpatrick et al., 2010).","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"We were intentionally lenient with our baselines: bilingual information by projecting POS tags directly across alignments in the parallel data.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"For a sentence x and a state sequence z, a first order Markov model defines a distribution: (9) where Val(X) corresponds to the entire vocabulary.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Mandarin exhibits several such processes, including A-not-A question formation, il\u00c2\u00ad lustrated in (3a), and adverbial reduplication, illustrated in (3b): 3.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"The Penn Arabic Treebank (ATB) syntactic guidelines (Maamouri et al., 2004) were purposefully borrowed without major modification from English (Marcus et al., 1993).","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"Table 1: Syntactic Seeding Heuristics BABAR\u00e2\u0080\u0099s reliable case resolution heuristics produced a substantial set of anaphor\/antecedent resolutions that will be the training data used to learn contextual role knowledge.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"D o m ai n # of ph ras es t o t a l p h r a s e s ac cu ra cy C C 7 o r m o r e 1 0 5 8 7 . 6 % 6 o r l e s s 1 0 6 6 7 . 0 % P C 7 o r m o r e 3 5 9 9 9 . 2 % 6 o r l e s s 2 5 5 6 5 . 1 % Table 1.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"The corpus-based statistical parsing community has many fast and accurate automated parsing systems, including systems produced by Collins (1997), Charniak (1997) and Ratnaparkhi (1997).","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"The effect of UNIQ is very small on both data sets.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"We hypothesize that modeling morphological information will greatly constrain the set of possible tags, thereby further refining the representation of the tag lexicon.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"Using these states, we define partial hypothesis extensions, which are of the following type: (S0;C n fjg; j0) !","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"In IdentiFinder, there is a priority in the feature assignment, such that if one feature is used for a token, another feature lower in priority will not be used.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"An analysis of nouns that occur in both the singular and the plural in our database reveals that there is indeed a slight but significant positive correlation-R2 = 0.20, p < 0.005; see Figure 6.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"This paper proposes a simple and effective tagging method that directly models tag sparsity and other distributional properties of valid POS tag assignments.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Note that the good classes JADE, GOLD and GRASS have lower costs than the bad classes SICKNESS, DEATH and RAT, as desired, so the trend observed for the results of this method is in the right direction.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"This has solutions: where pI(s|t) is derived from the IN corpus using relative-frequency estimates, and po(s|t) is an instance-weighted model derived from the OUT corpus.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"of Tokens No.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"Quite often, though, these directives fulfill the goal of increasing annotator agreement without in fact settling the theoretical question; i.e., the directives are clear but not always very well motivated.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"By contrast, BerkeleyLM\u2019s hash and compressed variants will return incorrect results based on an n \u22121-gram.","label":0},{"source_text":"There is no global pruning.","target_text":"The above auxiliary quantity satisfies the following recursive DP equation: Qe0 (e; C; j) = Initial Skip Verb Final 1.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Two of the Mainlanders also cluster close together but, interestingly, not particularly close to the Taiwan speakers; the third Mainlander is much more similar to the Taiwan speakers.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"We create equivalence classes for verb, noun, and adjective POS categories.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"The data structure was populated with 64-bit integers sampled uniformly without replacement.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"0 68.4 48.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"The errors shown are from the Berkeley parser output, but they are representative of the other two parsing models.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"7 68.3 56.","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"All commentaries have been annotated with rhetorical structure, using RSTTool4 and the definitions of discourse relations provided by Rhetorical Structure Theory (Mann, Thompson 1988).","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"Given an anaphor, BABAR identifies the caseframe that would extract it from its sentence.","label":0},{"source_text":"Here we present two algorithms.","target_text":".","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"Therefore we opted to use the much 7 The DempsterShafer theory assumes that one of the hypotheses in \u00ce\u00b8 is correct, so eliminating all of the hypotheses violates this assumption.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"While size of the resulting transducers may seem daunting-the segmenter described here, as it is used in the Bell Labs Mandarin TTS system has about 32,000 states and 209,000 arcs-recent work on minimization of weighted machines and transducers (cf.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"Thus at each iteration the algorithm is forced to pick features for the location, person and organization in turn for the classifier being trained.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"These enable much larger models in memory, compensating for lost accuracy.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"For the experiments in this paper we made a couple of additional modifications to the CoBoost algorithm.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"(2010) reports the best unsupervised results for English.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"This combination generalizes (2) and (3): we use either at = a to obtain a fixed-weight linear combination, or at = cI(t)\/(cI(t) + 0) to obtain a MAP combination.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"It was also proposed to allow annotators to skip sentences that they are unable to judge.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"For alif with hamza, normalization can be seen as another level of devocalization.","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"This leads to the best reported performance for robust non-projective parsing of Czech.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"The probabilities are incorporated into the DempsterShafer model using Equation 1.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"Formally, the DempsterShafer theory defines a probability density function m(S), where S is a set of hypotheses.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"In our experiment, we set the threshold of the TF\/ITF score empirically using a small development corpus; a finer adjustment of the threshold could reduce the number of such keywords.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"The BLEU score has been shown to correlate well with human judgement, when statistical machine translation systems are compared (Doddington, 2002; Przybocki, 2004; Li, 2005).","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":".t The path set of tree sets at level k +1 have the complexity of the string language of level k. The independence of paths in a tree set appears to be an important property.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"(See Sproat and Shih 1995.)","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"A name already mentioned previously in a document may appear in abbreviated form when it is mentioned again later.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"(2009), who also incorporate a sparsity constraint, but does via altering the model objective using posterior regularization.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"Our experiments all concern the analytical annotation, and the first experiment is based only on the training part.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"2.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"3.1 Maximum Entropy.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"In particular, the decision to represent arguments in verb- initial clauses as VP internal makes VSO and VOS configurations difficult to distinguish.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"The Leaf Ancestor metric measures the cost of transforming guess trees to the reference (Sampson and Babarczy, 2003).","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"After tagging a large corpus with an automatic NE tagger, the method tries to find sets of paraphrases automatically without being given a seed phrase or any kinds of cue.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"The first author is supported by a National Defense Science and Engineering Graduate (NDSEG) fellowship.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"This FSA I can be segmented into words by composing Id(I) with D*, to form the WFST shown in Figure 2(c), then selecting the best path through this WFST to produce the WFST in Figure 2(d).","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"They are also labelled for their topicality (yes \/ no), and this annotation is accompanied by a confidence value assigned by the annotator (since it is a more subjective matter).","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"For all languages we do not make use of a tagging dictionary.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"We will refer to our system as MENERGI (Maximum Entropy Named Entity Recognizer using Global Information).","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Another question that remains unanswered is to what extent the linguistic information he considers can be handled-or at least approximated-by finite-state language models, and therefore could be directly interfaced with the segmentation model that we have presented in this paper.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"Consider for a set of constituents the isolated constituent precision parser metric, the portion of isolated constituents that are correctly hypothesized.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"3.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"SRILM inefficiently stores 64-bit pointers.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"2.6 Co-reference.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"Instead, we want to apply an inverse transformation to recover the underlying (nonprojective) dependency graph.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"One striking example is the error reduction for Spanish, which reduces error by 36.5% and 24.7% for the best and median settings respectively.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"8 We use head-finding rules specified by a native speaker.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Further, the probing hash table does only one random lookup per query, explaining why it is faster on large data.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Formally, for a lexicon L and segments I \u00e2\u0088\u0088 L, O \u00e2\u0088\u0088\/ L, each word automaton accepts the language I\u00e2\u0088\u0097(O + I)I\u00e2\u0088\u0097.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"An ATM may be thought of as spawning independent processes for each applicable move.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"For each domain, phrases which contain the same keyword are gathered to build a set of phrases (Step 3).","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"The segmenter handles the grouping of hanzi into words and outputs word pronunciations, with default pronunciations for hanzi it cannot group; we focus here primarily on the system's ability to segment text appropriately (rather than on its pronunciation abilities).","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"It is well-known that constituency parsing models designed for English often do not generalize easily to other languages and treebanks.1 Explanations for this phenomenon have included the relative informativeness of lexicalization (Dubey and Keller, 2003; Arun and Keller, 2005), insensitivity to morphology (Cowan and Collins, 2005; Tsarfaty and Sima\u00e2\u0080\u0099an, 2008), and the effect of variable word order (Collins et al., 1999).","label":1},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"In addition to the automatic methods, AG, GR, and ST, just discussed, we also added to the plot the values for the current algorithm using only dictionary entries (i.e., no productively derived words or names).","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Also there are cases where one of the two NEs belong to a phrase outside of the relation.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"However, until such standards are universally adopted in evaluating Chinese segmenters, claims about performance in terms of simple measures like percent correct should be taken with a grain of salt; see, again, Wu and Fung (1994) for further arguments supporting this conclusion.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"As is standard, we report the greedy one-to-one (Haghighi and Klein, 2006) and the many-to-one token-level accuracy obtained from mapping model states to gold POS tags.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"We further thank Dr. J.-S.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"We include a list of per-category results for selected phrasal labels, POS tags, and dependencies in Table 8.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"Limitations There are several limitations in the methods.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"In the case of, the most common usage is as an adverb with the pronunciation jiangl, so that variant is assigned the estimated cost of 5.98, and a high cost is assigned to nominal usage with the pronunciation jiang4.","label":0},{"source_text":"Here we present two algorithms.","target_text":"The second algorithm builds on a boosting algorithm called AdaBoost (Freund and Schapire 97; Schapire and Singer 98).","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"For monolingual treebank data we relied on the CoNLL-X and CoNLL-2007 shared tasks on dependency parsing (Buchholz and Marsi, 2006; Nivre et al., 2007).","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"The entries in such a lexicon may be thought of as meaningful surface segments paired up with their PoS tags li = (si, pi), but note that a surface segment s need not be a space-delimited token.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Section 4 presents experimen tal results on two corpora: the MUC4 terrorism corpus, and Reuters texts about natural disasters.","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"Table 2 shows the features used in the current version of the parser.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"For \u00c3\u0086 = 1, a new target language word is generated using the trigram language model p(eje0; e00).","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"m(S) represents the belief that the correct hypothesis is included in S. The model assumes that evidence also arrives as a probability density function (pdf) over sets of hypotheses.6 Integrating new evidence into the existing model is therefore simply a matter of defining a function to merge pdfs, one representing the current belief system and one representing the beliefs of the new evidence.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"The final block in table 2 shows models trained on feature subsets and on the SVM feature described in 3.4.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"It is chosen such that the decisions it made in including or excluding constituents are most probable under the models for all of the parsers.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Participants and other volunteers contributed about 180 hours of labor in the manual evaluation.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"For each pair we also record the context, i.e. the phrase between the two NEs (Step1).","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"To formalize the approach, we introduce four verbgroup states S: Initial (I): A contiguous, initial block of source positions is covered.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"In our particular case the majority requires the agreement of only two parsers because we have only three.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"A named entity recognizer (NER) is useful in many NLP applications such as information extraction, question answering, etc. On its own, a NER can also provide users who are looking for person or organization names with quick information.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"First, we parsed the training corpus, collected all the noun phrases, and looked up each head noun in WordNet (Miller, 1990).","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"It is sometimes claimed that one of the advantages of dependency grammar over approaches based on constituency is that it allows a more adequate treatment of languages with variable word order, where discontinuous syntactic constructions are more common than in languages like English (Mel\u2019\u02c7cuk, 1988; Covington, 1990).","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"So, who won the competition?","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"Then, it can be verified that We can now derive the CoBoost algorithm as a means of minimizing Zco.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"For example, if CFLex determines that the log- likelihood statistic for the co-occurrence of a particular noun and caseframe corresponds to the 90% confidence level, then CFLex returns .90 as its belief that the anaphor and candidate are coreferent.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"Our analysis and comparison focuses primarily on the one-to-one accuracy since it is a stricter metric than many-to-one accuracy, but also report many-to-one for completeness.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"This withdrawal by the treasury secretary is understandable, though.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"The contextual role knowledge that BABAR uses for coreference resolution is derived from this caseframe data.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"Confidence Interval: To estimate confidence intervals for the average mean scores for the systems, we use standard significance testing.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"There is a sizable literature on Chinese word segmentation: recent reviews include Wang, Su, and Mo (1990) and Wu and Tseng (1993).","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"Statistics for all data sets are shown in Table 2.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"Final (F): The rest of the sentence is processed monotonically taking account of the already covered positions.","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"Special thanks to Jan Haji\u02c7c and Matthias Trautner Kromann for assistance with the Czech and Danish data, respectively, and to Jan Haji\u02c7c, Tom\u00b4a\u02c7s Holan, Dan Zeman and three anonymous reviewers for valuable comments on a preliminary version of the paper.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"+ cost(unseen(fm, as desired.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"We now describe the CoBoost algorithm for the named entity problem.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"Hence we decided to select ten commentaries to form a \u00e2\u0080\u0098core corpus\u00e2\u0080\u0099, for which the entire range of annotation levels was realized, so that experiments with multi-level querying could commence.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Moses keeps language models and many other resources in static variables, so these are still resident in memory.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"As mentioned above, it is not obvious how to apply Daum\u00b4e\u2019s approach to multinomials, which do not have a mechanism for combining split features.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Because the Bikel parser has been parameter- ized for Arabic by the LDC, we do not change the default model settings.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"This smooth guarantees that there are no zeroes estimated.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Our initial experimentation with the evaluation tool showed that this is often too overwhelming.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"If is one of Monday, Tuesday, . . .","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"For queries, we uniformly sampled 10 million hits and 10 million misses.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"In our experiments we set the parameter values randomly, and then ran EM to convergence.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"For instance, by altering the emission distribution parameters, Johnson (2007) encourages the model to put most of the probability mass on few tags.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"This overview is illustrated in Figure 1.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Case and Zone: If the token starts with a capital letter (initCaps), then an additional feature (init- Caps, zone) is set to 1.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"We introduce several new ideas.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"The motivation for our more informal approach was the intuition that there are so many open problems in rhetorical analysis (and more so for German than for English; see below) that the main task is qualitative investigation, whereas rigorous quantitative analyses should be performed at a later stage.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"The segmenter will give both analyses :1 cai2 neng2 'just be able,' and ?]cai2neng2 'talent,' but the latter analysis is preferred since splitting these two morphemes is generally more costly than grouping them.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"The method being described-henceforth ST..","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"In this case, we have no finite-state restrictions for the search space.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"We can do that . IbmS: Yes, wonderful.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"Unsupervised learning approaches appear to be a natural solution to this problem, as they require only unannotated text for training models.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"The development of automatic scoring methods is an open field of research.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"First, a non-anaphoric NP classifier identifies definite noun phrases that are existential, using both syntactic rules and our learned existential NP recognizer (Bean and Riloff, 1999), and removes them from the resolution process.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"Automatic scores are computed on a larger tested than manual scores (3064 sentences vs. 300\u2013400 sentences). collected manual judgements, we do not necessarily have the same sentence judged for both systems (judges evaluate 5 systems out of the 8\u201310 participating systems).","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"(student+plural) 'students,' which is derived by the affixation of the plural affix f, menD to the nounxue2shengl.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"A spelling rule might be a simple look-up for the string (e.g., a rule that Honduras is a location) or a rule that looks at words within a string (e.g., a rule that any string containing Mr. is a person).","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"For each language and setting, we report one-to-one (11) and many- to-one (m-1) accuracies.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"Different sentence structure and rich target language morphology are two reasons for this.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"This model admits a simple Gibbs sampling algorithm where the number of latent variables is proportional to the number of word types, rather than the size of a corpus as for a standard HMM sampler (Johnson, 2007).","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"The importance of dictionaries in NERs has been investigated in the literature (Mikheev et al., 1999).","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"ogy (Koskenniemi 1983; Antworth 1990; Tzoukermann and Liberman 1990; Karttunen, Kaplan, and Zaenen 1992; Sproat 1992); we represent the fact that ir, attaches to nouns by allowing t:-transitions from the final states of all noun entries, to the initial state of the sub-WFST representing f,.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"For example: McCann initiated a new global system.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Figure 4 Input lattice (top) and two segmentations (bottom) of the sentence 'How do you say octopus in Japanese?'.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"This combination generalizes (2) and (3): we use either at = a to obtain a fixed-weight linear combination, or at = cI(t)\/(cI(t) + 0) to obtain a MAP combination.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"The government has to make a decision, and do it quickly.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"to explore how well we can induce POS tags using only the one-tag-per-word constraint.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"72 77.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Table 1 shows results of the benchmark.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"For instance, on Spanish, the absolute gap on median performance is 10%.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"encodes the one tag per word constraint and is uni form over type-level tag assignments.","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"This is true of the widely used link grammar parser for English (Sleator and Temperley, 1993), which uses a dependency grammar of sorts, the probabilistic dependency parser of Eisner (1996), and more recently proposed deterministic dependency parsers (Yamada and Matsumoto, 2003; Nivre et al., 2004).","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Most computational models for coreference resolution rely on properties of the anaphor and candidate antecedent, such as lexical matching, grammatical and syntactic features, semantic agreement, and positional information.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"Besides information structure, the second main goal is to enhance current models of rhetorical structure.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"The type of alignment we have considered so far requires the same length for source and target sentence, i.e. I = J. Evidently, this is an unrealistic assumption, therefore we extend the concept of inverted alignments as follows: When adding a new position to the coverage set C, we might generate either \u00c3\u0086 = 0 or \u00c3\u0086 = 1 new target words.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"We have already mentioned the closely related work by Matsoukas et al (2009) on discriminative corpus weighting, and Jiang and Zhai (2007) on (nondiscriminative) instance weighting.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"But diacritics are not present in unvocalized text, which is the standard form of, e.g., news media documents.3 VBD she added VP PUNC S VP VBP NP ...","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"The toplevel weights are trained to maximize a metric such as BLEU on a small development set of approximately 1000 sentence pairs.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"Skipped (K): The translation of up to one word may be postponed . Verb (V): The translation of up to two words may be anticipated.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"Acknowledgments We thank Steven Bethard, Evan Rosen, and Karen Shiells for material contributions to this work.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"While we had up to 11 submissions for a translation direction, we did decide against presenting all 11 system outputs to the human judge.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"An important aspect of the DempsterShafer model is that it operates on sets of hypotheses.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Further, it needs extra pointers in the trie, increasing model size by 40%.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"We use the default inference parameters.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"5 67.3 55.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"There has been a lot of research on such lexical relations, along with the creation of resources such as WordNet.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"There has also been some work on adapting the word alignment model prior to phrase extraction (Civera and Juan, 2007; Wu et al., 2005), and on dynamically choosing a dev set (Xu et al., 2007).","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"In particular, it may not be possible to learn functions fi (x f2(x2,t) for i = m + 1...n: either because there is some noise in the data, or because it is just not realistic to expect to learn perfect classifiers given the features used for representation.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"10 Here we use the Good-Turing estimate (Baayen 1989; Church and Gale 1991), whereby the aggregate probability of previously unseen instances of a construction is estimated as ni\/N, where N is the total number of observed tokens and n1 is the number of types observed only once.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"Each knowledge source then assigns a probability estimate to each candidate, which represents its belief that the candidate is the antecedent for the anaphor.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"Many human evaluation metrics have been proposed.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"Supervised learning approaches have advanced the state-of-the-art on a variety of tasks in natural language processing, resulting in highly accurate systems.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Making the reasonable assumption that similar information is relevant for solving these problems in Chinese, it follows that a prerequisite for intonation-boundary assignment and prominence assignment is word segmentation.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"4.1 Corpora.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"4.2 A Sample Segmentation Using Only Dictionary Words Figure 4 shows two possible paths from the lattice of possible analyses of the input sentence B X:\u00c2\u00a5 .:.S:P:l 'How do you say octopus in Japanese?' previously shown in Figure 1.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"TPT has theoretically better locality because it stores ngrams near their suffixes, thereby placing reads for a single query in the same or adjacent pages.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"From the discussion so far it is clear that a number of formalisms involve some type of context-free rewriting (they have derivation trees that are local sets).","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Again, the idea is that having a picture of syntax, co-reference, and sentence-internal information structure at one\u00e2\u0080\u0099s disposal should aid in finding models of discourse structure that are more explanatory and can be empirically supported.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"For sorted lookup, we compare interpolation search, standard C++ binary search, and standard C++ set based on red-black trees.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"Borth- wick (1999) successfully made use of other hand- coded systems as input for his MENE system, and achieved excellent results.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"Experiments are presented in section 4.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"Since the inclusion of out-ofdomain test data was a very late decision, the participants were not informed of this.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"Section 5 covers relevant previous work on SMT adaptation, and section 6 concludes.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Alon Lavie advised on this work.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"The correct resolution in sentence (c) depends on knowledge that kidnappers frequently blindfold their victims.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"The rhetorical structure annotations of PCC have all been converted to URML.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"A generative model was applied (similar to naive Bayes) with the three labels as hidden vanables on unlabeled examples, and observed variables on (seed) labeled examples.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"The results, along with the total number of phrases, are shown in Table 1.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"We tried two versions of our graph-based approach: feature after the first stage of label propagation (Eq.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"'Malaysia.'","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"We carried out translation experiments in two different settings.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"4.1 The Task and the Corpus.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":":zhong1 : 0.0 tjl :huo2 :0.0 (R:spub:\/ic of Ch:ina) + .,_,...I : jlong4 :0.0 (mUifaty genG181) 0 \u00c2\u00a3: _NC: 40.0 Figure 3 Partial Chinese Lexicon (NC = noun; NP = proper noun).c=- - I \u00e2\u0080\u00a2=- :il: .;ss:;zhangt \u00e2\u0080\u00a2 '-:.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"(1) CEO of McCann . . .","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"But it conflates the coordinating and discourse separator functions of wa (<..4.b \u00ef\u00bf\u00bd \u00ef\u00bf\u00bd) into one analysis: conjunction(Table 3).","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"evaluated to account for the same fraction of the data.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"When keys are longer than 64 bits, we conserve space by replacing the keys with their 64-bit hashes.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"This is not ideal for some applications, however.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"As you can see in the figure, the accuracy for the domain is quite high except for the \u00e2\u0080\u009cagree\u00e2\u0080\u009d set, which contains various expressions representing different relationships for an IE application.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"We do not attempt to identify the types of relationships that are found.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"See Figure 3 for a screenshot of the evaluation tool.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"For natural disasters, BABAR generated 20,479 resolutions: 11,652 from lexical seeding and 8,827 from syntactic seeding.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"This decreases the statistical significance of our results compared to those studies.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"BABAR uses unsupervised learning to acquire this knowledge from plain text without the need for annotated training data.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"Our analysis identifies three key factors driving our performance gain: 1) selecting a model structure which directly encodes tag sparsity, 2) a type-level prior on tag assignments, and 3) a straightforward na\u00c2\u00a8\u00c4\u00b1veBayes approach to incorporate features.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"Gazdar (1985) considers a restriction of IG's in which no more than one nonterminal on the right-hand-side of a production can inherit the stack from the left-hand-side.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"The secondary classifier in (Borthwick, 1999) uses information not just from the current article, but also from the whole test corpus, with an additional feature that indicates if the information comes from the same document or from another document.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"The rewrite rules and the definition of the composition operations may be stored in the finite state control since G uses a finite number of them.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"While many systems had similar performance, the results offer interesting insights, especially about the relative performance of statistical and rule-based systems.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"The Potsdam Commentary Corpus","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Sheffield Guthrie and Hepple (2010) explore several randomized compression techniques, but did not release code.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"At present, the \u00e2\u0080\u0098Potsdam Commentary Corpus\u00e2\u0080\u0099 (henceforth \u00e2\u0080\u0098PCC\u00e2\u0080\u0099 for short) consists of 170 commentaries from Ma\u00c2\u00a8rkische Allgemeine Zeitung, a German regional daily.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"The major problem for all segmentation systems remains the coverage afforded by the dictionary and the lexical rules used to augment the dictionary to deal with unseen words.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"Realizing gains in practice can be challenging, however, particularly when the target domain is distant from the background data.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"The dictionary sizes reported in the literature range from 17,000 to 125,000 entries, and it seems reasonable to assume that the coverage of the base dictionary constitutes a major factor in the performance of the various approaches, possibly more important than the particular set of methods used in the segmentation.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Nicola Bertoldi and Marcello Federico assisted with IRSTLM.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"The resulting algorithm has a complexity of O(n!).","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"We confirm the finding by Callison-Burch et al. (2006) that the rule-based system of Systran is not adequately appreciated by BLEU.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"Our second contribution is to apply instance weighting at the level of phrase pairs.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"13.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"(1997; 1999) do have some statistics that show how IdentiFinder performs when the training data is reduced.","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"However, while Dienes and Dubey recognize empty categories in a pre-processing step and only let the parser find their antecedents, we use the parser both to detect dislocated dependents and to predict either the type or the location of their syntactic head (or both) and use post-processing only to transform the graph in accordance with the parser\u2019s analysis.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"68 96.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"(2) Description Source Location Names http:\/\/www.timeanddate.com http:\/\/www.cityguide.travel-guides.com http:\/\/www.worldtravelguide.net Corporate Names http:\/\/www.fmlx.com Person First Names http:\/\/www.census.gov\/genealogy\/names Person Last Names Table 2: Sources of Dictionaries The McCann family . . .","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"Applying the function PROJECTIVIZE to the graph in Figure 1 yields the graph in Figure 2, where the problematic arc pointing to Z has been lifted from the original head jedna to the ancestor je.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"7 Unlike Dickinson (2005), we strip traces and only con-.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"The improvement is due to the cost of bit-level reads and avoiding reads that may fall in different virtual memory pages.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"02 99.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"To conserve memory at the expense of accuracy, values may be quantized using q bits per probability and r bits per backoff6.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"input token, the segmentation is then performed deterministically given the 1-best analysis.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"Instead, we condition on the type-level tag assignments T . Specifically, let St = {i|Ti = t} denote the indices of theword types which have been assigned tag t accord ing to the tag assignments T . Then \u00ce\u00b8t is drawn from DIRICHLET(\u00ce\u00b1, St), a symmetric Dirichlet which only places mass on word types indicated by St. This ensures that each word will only be assigned a single tag at inference time (see Section 4).","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"What is important and is not expressed by the notation is the so-called coverage constraint: each source position j should be 'hit' exactly once by the path of the inverted alignment bI 1 = b1:::bi:::bI . Using the inverted alignments in the maximum approximation, we obtain as search criterion: max I (p(JjI) max eI 1 ( I Yi=1 p(eijei\u00f4\u0080\u0080\u00801 i\u00f4\u0080\u0080\u00802) max bI 1 I Yi=1 [p(bijbi\u00f4\u0080\u0080\u00801; I; J) p(fbi jei)])) = = max I (p(JjI) max eI 1;bI 1 ( I Yi=1 p(eijei\u00f4\u0080\u0080\u00801 i\u00f4\u0080\u0080\u00802) p(bijbi\u00f4\u0080\u0080\u00801; I; J) p(fbi jei)])); where the two products over i have been merged into a single product over i. p(eijei\u00f4\u0080\u0080\u00801 i\u00f4\u0080\u0080\u00802) is the trigram language model probability.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"These results are promising and there are several avenues for improving on these results.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"\u00ce\u00b2 is the shared hyperparameter for the tag assignment prior and word feature multinomials.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"We represent all morphological analyses of a given utterance using a lattice structure.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"Table 5: Individual Performance of KSs for Disasters (e.g., \u00e2\u0080\u009cthe mayor\u00e2\u0080\u009d vs. \u00e2\u0080\u009cthe journalist\u00e2\u0080\u009d).","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"This paper, however, provides a comprehensive overview of the data collection effort and its current state.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"This design leads to a significant reduction in the computational complexity of training and inference.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"In the case of, the most common usage is as an adverb with the pronunciation jiangl, so that variant is assigned the estimated cost of 5.98, and a high cost is assigned to nominal usage with the pronunciation jiang4.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"Here, we process only full-form words within the translation procedure.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Better Arabic Parsing: Baselines, Evaluations, and Analysis","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"We tabulate this increase in Table 3.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Each word is terminated by an arc that represents the transduction between f and the part of speech of that word, weighted with an estimated cost for that word.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"(2009).","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"It also incorporates the Good-Turing method (Baayen 1989; Church and Gale 1991) in estimating the likelihoods of previously unseen con\u00c2\u00ad structions, including morphological derivatives and personal names.","label":1},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"2.1 Reliable Case Resolutions.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"raphy: A ren2 'person' is a fairly uncontroversial case of a monographemic word, and rplil zhong1guo2 (middle country) 'China' a fairly uncontroversial case of a di\u00c2\u00ad graphernic word.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"However, for multinomial models like our LMs and TMs, there is a one to one correspondence between instances and features, eg the correspondence between a phrase pair (s, t) and its conditional multinomial probability p(s1t).","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"We used C = 1.0 as the L2 regularization constant in (Eq.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"The number of top-ranked pairs to retain is chosen to optimize dev-set BLEU score.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"The developers suggested some changes, such as building the model from scratch with IRSTLM, but these did not resolve the problem.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"A few annotators suggested to break up long sentences into clauses and evaluate these separately.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"The first setting uses the European Medicines Agency (EMEA) corpus (Tiedemann, 2009) as IN, and the Europarl (EP) corpus (www.statmt.org\/europarl) as OUT, for English\/French translation in both directions.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Email: rlls@bell-labs.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"Label propagation can then be used to transfer the labels to the peripheral foreign vertices (i.e. the ones adjacent to the English vertices) first, and then among all of the foreign vertices (\u00a74).","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Our TRIE implements the popular reverse trie, in which the last word of an n-gram is looked up first, as do SRILM, IRSTLM\u2019s inverted variant, and BerkeleyLM except for the scrolling variant.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"During coreference resolution, BABAR checks (1) whether the anaphor is among the lexical expectations for the caseframe that extracts the candidate antecedent, and (2) whether the candidate is among the lexical expectations for the caseframe that extracts the anaphor.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"In addition, this formulation results in a dramatic reduction in the number of model parameters thereby, enabling unusually rapid training.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"Se ma nti c (a) filters candidate if its semantic tags d o n \u00e2\u0080\u0099 t i n t e r s e c t w i t h t h o s e o f t h e a n a p h o r .","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"Log-linear combination (loglin) improves on this in all cases, and also beats the pure IN system.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Because of this, we retokenized and lowercased submitted output with our own tokenizer, which was also used to prepare the training and test data.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"What ought to be developed now is an annotation tool that can make use of the format, allow for underspecified annotations and visualize them accordingly.","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"Finally, the focus\/background partition is annotated, together with the focus question that elicits the corresponding answer.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"This offers the well-known advantages for inter- changability, but it raises the question of how to query the corpus across levels of annotation.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"For the translation experiments, Eq. 2 is recursively evaluated.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Given the closeness of most systems and the wide over-lapping confidence intervals it is hard to make strong statements about the correlation between human judgements and automatic scoring methods such as BLEU.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Previous reports on Chinese segmentation have invariably cited performance either in terms of a single percent-correct score, or else a single precision-recall pair.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"Each derivation of a grammar can be represented by a generalized context-free derivation tree.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"The only supervision is in the form of 7 seed rules (namely, that New York, California and U.S. are locations; that any name containing Mr is a person; that any name containing Incorporated is an organization; and that I.B.M. and Microsoft are organizations).","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"The percentage scores on the axis labels represent the amount of variation in the data explained by the dimension in question.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"This is equivalent to the assumption used in probability estimation for na\u00efve Bayes classifiers, namely that the attribute values are conditionally independent when the target value is given.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"We call these N \u2212 1 words state.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"Once again we present both a non-parametric and a parametric technique for this task.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"The second term is a regularizer and encourages all type marginals to be uniform to the extent that is allowed by the first two terms (cf. maximum entropy principle).","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"\u00e2\u0080\u00a2 Anaphoric links: the annotator is asked to specify whether the anaphor is a repetition, partial repetition, pronoun, epithet (e.g., Andy Warhol \u00e2\u0080\u0093 the PopArt artist), or is-a (e.g., Andy Warhol was often hunted by photographers.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"To summarize, we provided: The performance of the baseline system is similar to the best submissions in last year\u2019s shared task.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Hence we decided to restrict ourselves to only information from the same document.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"The PCFG was trained from the same sections of the Penn Treebank as the other three parsers.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"00 76.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"Specifically, we (+FEATS) utilizes the tag prior as well as features (e.g., suffixes and orthographic features), discussed in Section 3, for the P (W |T , \u00cf\u0088) component.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"Supervised part-of-speech (POS) taggers, for example, approach the level of inter-annotator agreement (Shen et al., 2007, 97.3% accuracy for English).","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"While the Bootstrap method is slightly more sensitive, it is very much in line with the sign test on text blocks.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"6.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"Phrase-level granularity distinguishes our work from previous work by Matsoukas et al (2009), who weight sentences according to sub-corpus and genre membership.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"4.1 Local Features.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"Nonetheless, parse quality is much lower in the joint model because a lattice is effectively a long sentence.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"We aligned the texts at a sentence level across all four languages, resulting in 1064 sentence per language.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"It is difficult when IN and OUT are dissimilar, as they are in the cases we study.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"In figure 4, reverse relations are indicated by `*\u00e2\u0080\u0099 next to the frequency.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"We evaluated BABAR on two domains: terrorism and natural disasters.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"68 95.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"It may seem surprising to some readers that the interhuman agreement scores reported here are so low.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"3.1 Corpora.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"Our motivation for using DempsterShafer is that it provides a well-principled framework for combining evidence from multiple sources with respect to competing hypotheses.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"html 4 www.wagsoft.com\/RSTTool assigning rhetorical relations is a process loaded with ambiguity and, possibly, subjectivity.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"During testing, it is possible that the classifier produces a sequence of inadmissible classes (e.g., person begin followed by location unique).","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"The judgement of 4 in the first case will go to a vastly better system output than in the second case.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"88,962 (spelling,context) pairs were extracted as training data.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"The computing time is given in terms of CPU time per sentence (on a 450MHz PentiumIIIPC).","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"The evaluation framework for the shared task is similar to the one used in last year\u2019s shared task.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"The Stanford parser includes both the manually annotated grammar (\u00c2\u00a74) and an Arabic unknown word model with the following lexical features: 1.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"This can be repeated several times to collect a list of author \/ book title pairs and expressions.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Language models that contain wi must also contain prefixes wi for 1 G i G k. Therefore, when the model is queried for p(wnjwn\u22121 1 ) but the longest matching suffix is wnf , it may return state s(wn1) = wnf since no longer context will be found.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"The rhetorical structure annotations of PCC have all been converted to URML.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"In addition, this formulation results in a dramatic reduction in the number of model parameters thereby, enabling unusually rapid training.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"Specifically, we (+FEATS) utilizes the tag prior as well as features (e.g., suffixes and orthographic features), discussed in Section 3, for the P (W |T , \u00cf\u0088) component.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Along with IRSTLM and TPT, our binary format is memory mapped, meaning the file and in-memory representation are the same.","label":0},{"source_text":"There is no global pruning.","target_text":"Subsets C of increasing cardinality c are processed.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"8 57.3 +F EA TS be st me dia n 50.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Conversely, the lattice parser requires no linguistic resources and produces segmentations of comparable quality.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Their results are then compared with the results of an automatic segmenter.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"If we remove this sample from the evaluation, then the ATB type-level error rises to only 37.4% while the n-gram error rate increases to 6.24%.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"We are given a source string fJ 1 = f1:::fj :::fJ of length J, which is to be translated into a target string eI 1 = e1:::ei:::eI of length I. Among all possible target strings, we will choose the string with the highest probability: ^eI 1 = arg max eI 1 fPr(eI 1jfJ 1 )g = arg max eI 1 fPr(eI 1) Pr(fJ 1 jeI 1)g : (1) The argmax operation denotes the search problem, i.e. the generation of the output sentence in the target language.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"We are unaware of prior results for the Stanford parser.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"As an indication, in our core corpus, we found an average sentence length of 15.8 words and 1.8 verbs per sentence, whereas a randomly taken sample of ten commentaries from the national papers Su\u00c2\u00a8ddeutsche Zeitung and Frankfurter Allgemeine has 19.6 words and 2.1 verbs per sentence.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"Previous work has attempted to incorporate such constraints into token-level models via heavy-handed modifications to inference procedure and objective function (e.g., posterior regularization and ILP decoding) (Grac\u00c2\u00b8a et al., 2009; Ravi and Knight, 2009).","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Compared with SRILM, IRSTLM adds several features: lower memory consumption, a binary file format with memory mapping, caching to increase speed, and quantization.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"the number of permutations carried out for the word reordering is given.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"The effect of a second reference resolution classifier is not entirely the same as that of global features.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"We again adopt an approach where we alternate between two classifiers: one classifier is modified while the other remains fixed.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"(b) does the translation have the same meaning, including connotations?","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Purely statistical approaches have not been very popular, and so far as we are aware earlier work by Sproat and Shih (1990) is the only published instance of such an approach.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"For more on the participating systems, please refer to the respective system description in the proceedings of the workshop.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"Our work is closest to that of Yarowsky and Ngai (2001), but differs in two important ways.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"IRSTLM\u2019s quantized variant is the inspiration for our quantized variant.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"If 7' is an initial tree we do not include an address on the left-hand side.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"The number of judgements is additionally fragmented by our breakup of sentences into in-domain and out-of-domain.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"We maintain a separate array for each length n containing all n-gram entries sorted in suffix order.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Arguably this consists of about three phonological words.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Since the segmentation corresponds to the sequence of words that has the lowest summed unigram cost, the segmenter under discussion here is a zeroth-order model.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"Hence, we use the bootstrap resampling method described by Koehn (2004).","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"(f1; ;mg n fl1; l2g ; l) 4 (f1; ;m \u00f4\u0080\u0080\u0080 1g n fl1; l2; l3g ; l0) !","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Diacritics can also be used to specify grammatical relations such as case and gender.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"In MUC6 and MUC7, the named entity task is defined as finding the following classes of names: person, organization, location, date, time, money, and percent (Chinchor, 1998; Sundheim, 1995) Machine learning systems in MUC6 and MUC 7 achieved accuracy comparable to rule-based systems on the named entity task.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"The trie data structure is commonly used for language modeling.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"The main disadvantage of manual evaluation is that it is time-consuming and thus too expensive to do frequently.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"The inventions disclosed herein are the subject of a patent application owned by the University of Utah and licensed on an exclusive basis to Attensity Corporation.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"A morpheme, on the other hand, usually corresponds to a unique hanzi, though there are a few cases where variant forms are found.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Two annotators received training with the RST definitions and started the process with a first set of 10 texts, the results of which were intensively discussed and revised.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Email: cls@bell-labs.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"The most accurate characterization of Chinese writing is that it is morphosyllabic (DeFrancis 1984): each hanzi represents one morpheme lexically and semantically, and one syllable phonologi\u00c2\u00ad cally.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"Notice that the CC-domain is a special case.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"All commentaries have been tagged with part-of-speech information using Brants\u00e2\u0080\u0099 TnT1 tagger and the Stuttgart\/Tu\u00c2\u00a8bingen Tag Set automatic analysis was responsible for this decision.)","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"-1 means that an NP should be ruled out as a possible antecedent, and 0 means that the knowledge source remains neutral (i.e., it has no reason to believe that they cannot be coreferent).","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"We use MMAX for this annotation as well.","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"The first thing to note is that projectivizing helps in itself, even if no encoding is used, as seen from the fact that the projective baseline outperforms the non-projective training condition by more than half a percentage point on attachment score, although the gain is much smaller with respect to exact match.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"However, such an approach requires a number of hand-coded systems, which may not be available in languages other than English.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"One annotator suggested that this was the case for as much as 10% of our test sentences.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"\u00e2\u0088\u0097NP NP PP R) and \u00e2\u0088\u0097NP NP ADJP R) are both iDafa attachment.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"The additional morphological material in such cases appears after the stem and realizes the extended meaning.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"The last four columns in Table 3 show the distribution of nonprojective arcs with respect to the number of lifts required.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"A spelling rule might be a simple look-up for the string (e.g., a rule that Honduras is a location) or a rule that looks at words within a string (e.g., a rule that any string containing Mr. is a person).","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"The improvement is due to the cost of bit-level reads and avoiding reads that may fall in different virtual memory pages.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"We then use linguistic and annotation insights to develop a manually annotated grammar for Arabic (\u00c2\u00a74).","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"While the Bootstrap method is slightly more sensitive, it is very much in line with the sign test on text blocks.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"1 61.2 43.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"The tee in Figure 9a can be denoted by t1 i223 where tee substitution is used instead of concatenation.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"Here, we process only full-form words within the translation procedure.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"The accuracy is calculated as the ratio of the number of paraphrases to the total number of phrases in the set.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Approaches differ in the algorithms used for scoring and selecting the best path, as well as in the amount of contextual information used in the scoring process.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"We confirm the finding by Callison-Burch et al. (2006) that the rule-based system of Systran is not adequately appreciated by BLEU.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Besides the lack of a clear definition of what constitutes a correct segmentation for a given Chinese sentence, there is the more general issue that the test corpora used in these evaluations differ from system to system, so meaningful comparison between systems is rendered even more difficult.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"By using the output of a hand-coded system such as Proteus, MENE can improve its performance, and can even outperform IdentiFinder (Borthwick, 1999).","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"Our suspicion is that BLEU is very sensitive to jargon, to selecting exactly the right words, and not synonyms that human judges may appreciate as equally good.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"In order to create an IE system for a new domain, one has to spend a long time to create the knowledge.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"A derived structure will be mapped onto a sequence zi of substrings (not necessarily contiguous in the input), and the composition operations will be mapped onto functions that can defined as follows3. f((zi,\u2022 \u2022 \u2022 , zni), (m.,\u2022 \u2022 \u2022 ,Yn3)) = (Z1, \u2022 \u2022 \u2022 , Zn3) where each z, is the concatenation of strings from z,'s and yk's.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"Conversely, the lattice parser requires no linguistic resources and produces segmentations of comparable quality.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Participants were also provided with two sets of 2,000 sentences of parallel text to be used for system development and tuning.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"None of the parsers produce parses with crossing brackets, so none of them votes for both of the assumed constituents.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"The toplevel weights are trained to maximize a metric such as BLEU on a small development set of approximately 1000 sentence pairs.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"In the next section we present an alternative approach that builds two classifiers while attempting to satisfy the above constraints as much as possible.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"Pairwise comparison is done using the sign test.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"Our initial experimentation with the evaluation tool showed that this is often too overwhelming.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Note also that the costs currently used in the system are actually string costs, rather than word costs.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"We will describe the evaluation of such clusters in the next subsection.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"An error count of 0:0 is assigned to a perfect translation, and an error count of 1:0 is assigned to a semantically and syntactically wrong translation.","label":0},{"source_text":"Here we present two algorithms.","target_text":"2.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"This is similar to using the Linux MAP POPULATE flag that is our default loading mechanism.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"In Figure 4, we displayed the number of system comparisons, for which we concluded statistical significance.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"The second weakness is purely conceptual, and probably does not affect the per\u00c2\u00ad formance of the model.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"In Figure 4 we show an example of variation between the parsing models.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"Section 5 covers relevant previous work on SMT adaptation, and section 6 concludes.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"Clearly, explicitly modeling such a powerful constraint on tagging assignment has a potential to significantly improve the accuracy of an unsupervised part-of-speech tagger learned without a tagging dictionary.","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"The first experiment uses data from two dependency treebanks.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"07 95.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"We will report the evaluation results in the next subsection.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"This significantly underperforms log-linear combination.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"Similar behavior is observed when adding features.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"We contrasted formalisms such as CFG's, HG's, TAG's and MCTAG's, with formalisms such as IG's and unificational systems such as LFG's and FUG's.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"(1998) did make use of information from the whole document.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"For displaying and querying the annoated text, we make use of the Annis Linguistic Database developed in our group for a large research effort (\u00e2\u0080\u0098Sonderforschungsbereich\u00e2\u0080\u0099) revolving around 9 2.7 Information structure.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"logical rules, and personal names; the transitive closure of the resulting machine is then computed.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Variants of alif are inconsistently used in Arabic texts.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"To our knowledge, ours is the first analysis of this kind for Arabic parsing.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"7.96 5.55 1 l...................................................................................................................................................................................................J..","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"We respond to this on the one hand with a format for its underspecification (see 2.4) and on the other hand with an additional level of annotation that attends only to connectives and their scopes (see 2.5), which is intended as an intermediate step on the long road towards a systematic and objective treatment of rhetorical structure.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"It can be shown that the path set of the tree set generated by a TAG G is a context-free language.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"It was also proposed to allow annotators to skip sentences that they are unable to judge.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"(If fewer than n rules have Precision greater than pin, we 3Note that taking tlie top n most frequent rules already makes the method robut to low count events, hence we do not use smoothing, allowing low-count high-precision features to be chosen on later iterations. keep only those rules which exceed the precision threshold.) pm,n was fixed at 0.95 in all experiments in this paper.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"2 We now introduce a new algorithm for learning from unlabeled examples, which we will call DLCoTrain (DL stands for decision list, the term Cotrain is taken from (Blum and Mitchell 98)).","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"The particular classifier used depends upon the noun.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Another option is the closedsource data structures from Sheffield (Guthrie and Hepple, 2010).","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"This line of work has been motivated by empirical findings that the standard EM-learned unsupervised HMM does not exhibit sufficient word tag sparsity.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Whether a language even has orthographic words is largely dependent on the writing system used to represent the language (rather than the language itself); the notion \"orthographic word\" is not universal.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"To facilitate the comparison of our results to those reported by (Cohen and Smith, 2007) we use their data set in which 177 empty and \u201cmalformed\u201d7 were removed.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"We carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"A total of 13,976 phrases were grouped.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"For parsing, the most challenging form of ambiguity occurs at the discourse level.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"7 68.3 56.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"For novel texts, no lexicon that consists simply of a list of word entries will ever be entirely satisfactory, since the list will inevitably omit many constructions that should be considered words.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"To lower the barrier of entrance to the competition, we provided a complete baseline MT system, along with data resources.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"If there is a frequent multi-word sequence in a domain, we could use it as a keyword candidate.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Replacing this with an ranked evaluation seems to be more suitable.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Church and Hanks [1989]), and we have used lists of character pairs ranked by mutual information to expand our own dictionary.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"We construct a mapping from all the space-delimited tokens seen in the training sentences to their corresponding analyses.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"The Bikel GoldPOS configuration only supplies the gold POS tags; it does not force the parser to use them.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"Similar advances have been made in machine translation (Frederking and Nirenburg, 1994), speech recognition (Fiscus, 1997) and named entity recognition (Borthwick et al., 1998).","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"It is well know that language pairs such as EnglishGerman pose more challenges to machine translation systems than language pairs such as FrenchEnglish.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"We present KenLM, a library that implements two data structures for efficient language model queries, reducing both time and costs.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"The developers aimed to reduce memory consumption at the expense of time.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"For example, in Figure 3, we can see that the phrases in the \u00e2\u0080\u009cbuy\u00e2\u0080\u009d, \u00e2\u0080\u009cacquire\u00e2\u0080\u009d and \u00e2\u0080\u009cpurchase\u00e2\u0080\u009d sets are mostly paraphrases.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"A second restriction on the formalisms is that choices during the derivation are independent of the context in the derivation.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"The baseline alignment model does not permit that a source word is aligned to two or more target words, e.g. for the translation direction from German toEnglish, the German compound noun 'Zahnarztter min' causes problems, because it must be translated by the two target words dentist's appointment.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Similarly, there is no compelling evidence that either of the syllables of f.ifflll binllang2 'betelnut' represents a morpheme, since neither can occur in any context without the other: more likely fjfflll binllang2 is a disyllabic morpheme.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"Both parametric and non-parametric models are explored.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Our system does not currently make use of titles, but it would be straightforward to do so within the finite-state framework that we propose.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"In the cases where isolated constituent precision is larger than 0.5 the affected portion of the hypotheses is negligible.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Even with less training data, MENERGI outperforms Borthwick' s MENE + reference resolution (Borthwick, 1999).","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"The particular classifier used depends upon the noun.","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"We then evaluate the approach in two steps.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"In Table 5 we present results from small test cor\u00c2\u00ad pora for the productive affixes handled by the current version of the system; as with names, the segmentation of morphologically derived words is generally either right or wrong.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"Other approaches encode sparsity as a soft constraint.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"Because of this, we retokenized and lowercased submitted output with our own tokenizer, which was also used to prepare the training and test data.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"The ability to redistribute belief values across sets rather than individual hypotheses is key.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"The state-to-tag mapping is obtained from the best hyperparameter setting for 11 mapping shown in Table 3.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"The second stage links sets which involve the same pairs of individual NEs.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"The developer explained that the loading process requires extra memory that it then frees. eBased on the ratio to SRI\u2019s speed reported in Guthrie and Hepple (2010) under different conditions.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"BABAR achieved recall in the 4250% range for both domains, with 76% precision overall for terrorism and 87% precision for natural disasters.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Historically, Arabic grammar has identified two sentences types: those that begin with a nominal (\u00ef\u00bf\u00bd '.i \u00ef\u00bf\u00bdu _..","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"99 94.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"The test accuracy more or less asymptotes.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"Restrictions: Quasi-monotone Search The above search space is still too large to allow the translation of a medium length input sentence.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"KenLM: Faster and Smaller Language Model Queries","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"But we will show that the use of unlabeled data can drastically reduce the need for supervision.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"We consider the unsupervised POS induction problem without the use of a tagging dictionary.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"The 1st block contains the simple baselines from section 2.1.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":".., Tn ) T W \u00cf\u0086 E w : Token word seqs (obs) t : Token tag assigns (det by T ) PARAMETERS \u00cf\u0088 : Lexicon parameters \u00ce\u00b8 : Token word emission parameters \u00cf\u0086 : Token tag transition parameters \u00cf\u0086 \u00cf\u0086 t1 t2 \u00ce\u00b8 \u00ce\u00b8 w1 w2 K \u00cf\u0086 T tm O K \u00ce\u00b8 E wN m N N Figure 1: Graphical depiction of our model and summary of latent variables and parameters.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"For the experiments, we used four newswire corpora, the Los Angeles Times\/Washington Post, The New York Times, Reuters and the Wall Street Journal, all published in 1995.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"3.2 Reordering with IBM Style.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"Unfortunately, the best completely unsupervised English POS tagger (that does not make use of a tagging dictionary) reaches only 76.1% accuracy (Christodoulopoulos et al., 2010), making its practical usability questionable at best.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"As indicated in Figure 1(c), apart from this correct analysis, there is also the analysis taking B ri4 as a word (e.g., a common abbreviation for Japan), along with X:\u00c2\u00a5 wen2zhangl 'essay\/ and f!!.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"10.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"(1), Pr(eI 1) is the language model, which is a trigram language model in this case.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"By taking the ratio of matching n-grams to the total number of n-grams in the system output, we obtain the precision pn for each n-gram order n. These values for n-gram precision are combined into a BLEU score: The formula for the BLEU metric also includes a brevity penalty for too short output, which is based on the total number of words in the system output c and in the reference r. BLEU is sensitive to tokenization.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"Extract NE pair instances with contexts From the four years of newspaper corpus, we extracted 1.9 million pairs of NE instances.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"This is the first time that we organized a large-scale manual evaluation.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"Link phrases based on instance pairs Using NE instance pairs as a clue, we find links between sets of phrases.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"Denote the unthresholded classifiers after t \u2014 1 rounds by git\u20141 and assume that it is the turn for the first classifier to be updated while the second one is kept fixed.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Still, for both human and automatic rhetorical analysis, connectives are the most important source of surface information.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"First, we learn weights on individual phrase pairs rather than sentences.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"This annotation choice weakens splitIN.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"The second case involves existential noun phrases (Allen, 1995), which are noun phrases that uniquely specify an object or concept and therefore do not need a prior referent in the discourse.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Time starts when Moses is launched and therefore includes model loading time.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"(b) F.i'JJI!","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"In this paper, we offer broad insight into the underperformance of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"In contrast, our morphological probabilities are based on a unigram, lexeme-based model, and all other (local and non-local) contextual considerations are delegated to the PCFG.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"The standard measures for evaluating Penn Treebank parsing performance are precision and recall of the predicted constituents.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"selected; and that recall is defined to be the number of correct hits divided by the number of items that should have been selected.","label":0},{"source_text":"There is no global pruning.","target_text":"(f1; ;mg n fl1; l2; l3g ;m) German to English the monotonicity constraint is violated mainly with respect to the German verbgroup.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Morphologically derived words such as, xue2shengl+men0.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"To conserve memory at the expense of accuracy, values may be quantized using q bits per probability and r bits per backoff6.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"The kind of annotation work presented here would clearly benefit from the emergence of standard formats and tag sets, which could lead to sharable resources of larger size.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"Denote the unthresholded classifiers after t \u2014 1 rounds by git\u20141 and assume that it is the turn for the first classifier to be updated while the second one is kept fixed.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"We use a simple TF\/IDF method to measure the topicality of words.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"7 www.cis.upenn.edu\/\u00e2\u0088\u00bcpdtb\/ 8 www.eml-research.de\/english\/Research\/NLP\/ Downloads had to buy a new car.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"was done by the participants.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"(S2) The burglar fired the gun three times and fled.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"19 We note that it is not always clear in Wang, Li, and Chang's examples which segmented words.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"These results are promising and there are several avenues for improving on these results.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"We define the path set of a tree 1 as the set of strings that label a path from the root to frontier of 7.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Following this method, we repeatedly \u2014 say, 1000 times \u2014 sample sets of sentences from the output of each system, measure their BLEU score, and use these 1000 BLEU scores as basis for estimating a confidence interval.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"We evaluate the time and memory consumption of each data structure by computing perplexity on 4 billion tokens from the English Gigaword corpus (Parker et al., 2009).","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"Recent research has demonstrated that incorporating this sparsity constraint improves tagging accuracy.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"This approach is compared to another reordering scheme presented in (Berger et al., 1996).","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"The second setting uses the news-related subcorpora for the NIST09 MT Chinese to English evaluation8 as IN, and the remaining NIST parallel Chinese\/English corpora (UN, Hong Kong Laws, and Hong Kong Hansard) as OUT.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"In Figure 4, we displayed the number of system comparisons, for which we concluded statistical significance.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"Our results show that BABAR achieves good performance in both domains, and that the contextual role knowledge improves performance, especially on pronouns.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"We group the features used into feature groups.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"\u00e2\u0080\u009d, because Bush is the first word, the initial caps might be due to its position (as in \u00e2\u0080\u009cThey put a freeze on . . .","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"While RST (Mann, Thompson 1988) proposed that a single relation hold between adjacent text segments, SDRT (Asher, Lascarides 2003) maintains that multiple relations may hold simultaneously.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"First we present the non-parametric version of parser switching, similarity switching: The intuition for this technique is that we can measure a similarity between parses by counting the constituents they have in common.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"The developers suggested some changes, such as building the model from scratch with IRSTLM, but these did not resolve the problem.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"However, some caveats are in order in comparing this method (or any method) with other approaches to seg\u00c2\u00ad mentation reported in the literature.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"constitute names, since we have only their segmentation, not the actual classification of the segmented words.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"During training, we treat as observed the language word types W as well as the token-level corpus w. We utilize Gibbs sampling to approximate our collapsed model posterior: P (T ,t|W , w, \u00ce\u00b1, \u00ce\u00b2) \u00e2\u0088\u009d P (T , t, W , w|\u00ce\u00b1, \u00ce\u00b2) 0.7 0.6 0.5 0.4 0.3 English Danish Dutch Germany Portuguese Spanish Swedish = P (T , t, W , w, \u00cf\u0088, \u00ce\u00b8, \u00cf\u0086, w|\u00ce\u00b1, \u00ce\u00b2)d\u00cf\u0088d\u00ce\u00b8d\u00cf\u0086 Note that given tag assignments T , there is only one setting of token-level tags t which has mass in the above posterior.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"Our goal is to define a class of formal systems, and show that any member of this class will possess certain attractive properties.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"IdentiFinder ' 99' s results are considerably better than IdentiFinder ' 97' s. IdentiFinder' s performance in MUC7 is published in (Miller et al., 1998).","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"The state-to-tag mapping is obtained from the best hyperparameter setting for 11 mapping shown in Table 3.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"The correct resolution in sentence (c) depends on knowledge that kidnappers frequently blindfold their victims.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"(7), such as the likelihood function used in maximum-entropy problems and other generalized additive models (Lafferty 99).","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"If two nouns have mutually exclusive semantic classes, then they cannot be coreferent.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Interestingly, Chang et al. report 80.67% recall and 91.87% precision on an 11,000 word corpus: seemingly, our system finds as many names as their system, but with four times as many false hits.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"One possibility is to use n-grams based on mutual information.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"selected; and that recall is defined to be the number of correct hits divided by the number of items that should have been selected.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"In this section, we try to compare our results with those obtained by IdentiFinder ' 97 (Bikel et al., 1997), IdentiFinder ' 99 (Bikel et al., 1999), and MENE (Borthwick, 1999).","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"The choice of the particular newspaper was motivated by the fact that the language used in a regional daily is somewhat simpler than that of papers read nationwide.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"For parsing, this is a mistake, especially in the case of interrogatives.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"First, the training data for the parser is projectivized by applying a minimal number of lifting operations (Kahane et al., 1998) and encoding information about these lifts in arc labels.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"Notice that even though IG's and LFG's involve CFG-like productions, they are (linguistically) fundamentally different from CFG's because the composition operations need not be linear.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"In section 2 we introduce the graph transformation techniques used to projectivize and deprojectivize dependency graphs, and in section 3 we describe the data-driven dependency parser that is the core of our system.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"99 94.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"A minimal requirement for building a Chinese word segmenter is obviously a dictionary; furthermore, as has been argued persuasively by Fung and Wu (1994), one will perform much better at segmenting text by using a dictionary constructed with text of the same genre as the text to be segmented.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"3.2 Results.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"To resolve the anaphor, we survey the final belief values assigned to each candidate\u00e2\u0080\u0099s singleton set.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Non-empty buckets contain an entry belonging to them or to a preceding bucket where a conflict occurred.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"The focus of our work is on the use of contextual role knowledge for coreference resolution.","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"Using this encoding scheme, the arc from je to Z in Figure 2 would be assigned the label AuxP\u2191Sb (signifying an AuxP that has been lifted from a Sb).","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"KS Function Ge nde r filters candidate if gender doesn\u00e2\u0080\u0099t agree.","label":0},{"source_text":"There is no global pruning.","target_text":"2.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"We do not adapt the alignment procedure for generating the phrase table from which the TM distributions are derived.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":".., Tn ) T W \u00cf\u0086 E w : Token word seqs (obs) t : Token tag assigns (det by T ) PARAMETERS \u00cf\u0088 : Lexicon parameters \u00ce\u00b8 : Token word emission parameters \u00cf\u0086 : Token tag transition parameters \u00cf\u0086 \u00cf\u0086 t1 t2 \u00ce\u00b8 \u00ce\u00b8 w1 w2 K \u00cf\u0086 T tm O K \u00ce\u00b8 E wN m N N Figure 1: Graphical depiction of our model and summary of latent variables and parameters.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"For each co-occurrence relation (noun\/caseframe for CFLex, and caseframe\/caseframe for CFNet), BABAR computes its log-likelihood value and looks it up in the \u00cf\u00872 table to obtain a confidence level.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"The simplest version of the maximum matching algorithm effectively deals with ambiguity by ignoring it, since the method is guaranteed to produce only one segmentation.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"Unsupervised Learning of Contextual Role Knowledge for Coreference Resolution","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"MUC7 has also seen hybrids of statistical NERs and hand-coded systems (Mikheev et al., 1998; Borthwick, 1999), notably Mikheev' s system, which achieved the best performance of 93.39% on the official NE test data.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"This is especially true in the case of quotations\u00e2\u0080\u0094which are common in the ATB\u00e2\u0080\u0094where (1) will follow a verb like (2) (Figure 1).","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"This WFST is then summed with the WFST implementing the dictionary and morphological rules, and the transitive closure of the resulting transducer is computed; see Pereira, Riley, and Sproat (1994) for an explanation of the notion of summing WFSTs.12 Conceptual Improvements over Chang et al.'s Model.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"For the translation model Pr(fJ 1 jeI 1), we go on the assumption that each source word is aligned to exactly one target word.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"Then, token- level HMM emission parameters are drawn conditioned on these assignments such that each word is only allowed probability mass on a single assigned tag.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"3) A tight coupling with the speech recognizer output.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"One may argue with these efforts on normalization, and ultimately their value should be assessed by assessing their impact on inter-annotator agreement.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"For example, from the sentence \u00e2\u0080\u009cMr.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"can expect famous names like Zhou Enlai's to be in many dictionaries, but names such as :fi lf;f; shi2jil-lin2, the name of the second author of this paper, will not be found in any dictionary.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Let us notate the set of previously unseen, or novel, members of a category X as unseen(X); thus, novel members of the set of words derived in f, menO will be de\u00c2\u00ad noted unseen(f,).","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"The standard measures for evaluating Penn Treebank parsing performance are precision and recall of the predicted constituents.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"We used the TF\/ITF metric to identify keywords.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"However, we needed to restrict ourselves to these languages in order to be able to evaluate the performance of our approach.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"Consider the case where IX].","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"If a phrase does not contain any keywords, the phrase is discarded.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"3.1 Maximum Entropy.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"Both parametric and non-parametric models are explored.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"The confidence intervals are computed by bootstrap resampling for BLEU, and by standard significance testing for the manual scores, as described earlier in the paper.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"markBaseNP indicates these non-recursive nominal phrases.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"Adequacy (rank) Fluency (rank) BLEU (rank) upc-jmc (1-7) (1-8) (1-6) lcc (1-6) (1-7) (1-4) utd (1-7) (1-6) (2-7) upc-mr (1-8) (1-6) (1-7) nrc (1-7) (2-6) (8) ntt (1-8) (2-8) (1-7) cmu (3-7) (4-8) (2-7) rali (5-8) (3-9) (3-7) systran (9) (8-9) (10) upv (10) (10) (9) Spanish-English (In Domain) Adequacy (rank) Fluency (rank) BLEU (rank) upc-jmc (1-7) (1-6) (1-5) ntt (1-7) (1-8) (1-5) lcc (1-8) (2-8) (1-4) utd (1-8) (2-7) (1-5) nrc (2-8) (1-9) (6) upc-mr (1-8) (1-6) (7) uedin-birch (1-8) (2-10) (8) rali (3-9) (3-9) (2-5) upc-jg (7-9) (6-9) (9) upv (10) (9-10) (10) German-English (In Domain) Adequacy (rank) Fluency (rank) BLEU (rank) uedin-phi (1-2) (1) (1) lcc (2-7) (2-7) (2) nrc (2-7) (2-6) (5-7) utd (3-7) (2-8) (3-4) ntt (2-9) (2-8) (3-4) upc-mr (3-9) (6-9) (8) rali (4-9) (3-9) (5-7) upc-jmc (2-9) (3-9) (5-7) systran (3-9) (3-9) (10) upv (10) (10) (9) Figure 7: Evaluation of translation to English on in-domain test data 112 English-French (In Domain) Adequacy (rank) Fluency (rank) BLEU (rank) nrc (1-5) (1-5) (1-6) upc-mr (1-4) (1-5) (1-6) upc-jmc (1-6) (1-6) (1-5) systran (2-7) (1-6) (7) utd (3-7) (3-7) (3-6) rali (1-7) (2-7) (1-6) ntt (4-7) (4-7) (1-5) English-Spanish (In Domain) Adequacy (rank) Fluency (rank) BLEU (rank) ms (1-5) (1-7) (7-8) upc-mr (1-4) (1-5) (1-4) utd (1-5) (1-6) (1-4) nrc (2-7) (1-6) (5-6) ntt (3-7) (1-6) (1-4) upc-jmc (2-7) (2-7) (1-4) rali (5-8) (6-8) (5-6) uedin-birch (6-9) (6-10) (7-8) upc-jg (9) (8-10) (9) upv (9-10) (8-10) (10) English-German (In Domain) Adequacy (rank) Fluency (rank) BLEU (rank) upc-mr (1-3) (1-5) (3-5) ntt (1-5) (2-6) (1-3) upc-jmc (1-5) (1-4) (1-3) nrc (2-4) (1-5) (4-5) rali (3-6) (2-6) (1-4) systran (5-6) (3-6) (7) upv (7) (7) (6) Figure 8: Evaluation of translation from English on in-domain test data 113 French-English (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) upc-jmc (1-5) (1-8) (1-4) cmu (1-8) (1-9) (4-7) systran (1-8) (1-7) (9) lcc (1-9) (1-9) (1-5) upc-mr (2-8) (1-7) (1-3) utd (1-9) (1-8) (3-7) ntt (3-9) (1-9) (3-7) nrc (3-8) (3-9) (3-7) rali (4-9) (5-9) (8) upv (10) (10) (10) Spanish-English (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) upc-jmc (1-2) (1-6) (1-3) uedin-birch (1-7) (1-6) (5-8) nrc (2-8) (1-8) (5-7) ntt (2-7) (2-6) (3-4) upc-mr (2-8) (1-7) (5-8) lcc (4-9) (3-7) (1-4) utd (2-9) (2-8) (1-3) upc-jg (4-9) (7-9) (9) rali (4-9) (6-9) (6-8) upv (10) (10) (10) German-English (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) systran (1-4) (1-4) (7-9) uedin-phi (1-6) (1-7) (1) lcc (1-6) (1-7) (2-3) utd (2-7) (2-6) (4-6) ntt (1-9) (1-7) (3-5) nrc (3-8) (2-8) (7-8) upc-mr (4-8) (6-8) (4-6) upc-jmc (4-8) (3-9) (2-5) rali (8-9) (8-9) (8-9) upv (10) (10) (10) Figure 9: Evaluation of translation to English on out-of-domain test data 114 English-French (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) systran (1) (1) (1) upc-jmc (2-5) (2-4) (2-6) upc-mr (2-4) (2-4) (2-6) utd (2-6) (2-6) (7) rali (4-7) (5-7) (2-6) nrc (4-7) (4-7) (2-5) ntt (4-7) (4-7) (3-6) English-Spanish (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) upc-mr (1-3) (1-6) (1-2) ms (1-7) (1-8) (6-7) utd (2-6) (1-7) (3-5) nrc (1-6) (2-7) (3-5) upc-jmc (2-7) (1-6) (3-5) ntt (2-7) (1-7) (1-2) rali (6-8) (4-8) (6-8) uedin-birch (6-10) (5-9) (7-8) upc-jg (8-9) (9-10) (9) upv (9) (8-9) (10) English-German (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) systran (1) (1-2) (1-6) upc-mr (2-3) (1-3) (1-5) upc-jmc (2-3) (3-6) (1-6) rali (4-6) (4-6) (1-6) nrc (4-6) (2-6) (2-6) ntt (4-6) (3-5) (1-6) upv (7) (7) (7) Figure 10: Evaluation of translation from English on out-of-domain test data 115 French-English In domain Out of Domain Adequacy Adequacy 0.3 0.3 \u2022 0.2 0.2 0.1 0.1 -0.0 -0.0 -0.1 -0.1 -0.2 -0.2 -0.3 -0.3 -0.4 -0.4 -0.5 -0.5 -0.6 -0.6 -0.7 -0.7 \u2022upv -0.8 -0.8 21 22 23 24 25 26 27 28 29 30 31 15 16 17 18 19 20 21 22 \u2022upv \u2022systran upcntt \u2022 rali upc-jmc \u2022 cc Fluency 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 \u2022upv -0.5 \u2022systran \u2022upv upc -jmc \u2022 Fluency 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 \u2022 \u2022 \u2022 td t cc upc- \u2022 rali 21 22 23 24 25 26 27 28 29 30 31 15 16 17 18 19 20 21 22 Figure 11: Correlation between manual and automatic scores for French-English 116 Spanish-English Figure 12: Correlation between manual and automatic scores for Spanish-English -0.3 -0.4 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 \u2022upv -0.4 \u2022upv -0.3 In Domain \u2022upc-jg Adequacy 0.3 0.2 0.1 -0.0 -0.1 -0.2 Out of Domain \u2022upc-jmc \u2022nrc \u2022ntt Adequacy upc-jmc \u2022 \u2022 \u2022lcc \u2022 rali \u2022 \u2022rali -0.7 -0.5 -0.6 \u2022upv 23 24 25 26 27 28 29 30 31 32 19 20 21 22 23 24 25 26 27 28 \u2022 \u2022rali Fluency 0.2 0.1 -0.0 -0.1 -0.2 ntt \u2022 upc-mr \u2022lcc \u2022utd \u2022upc-jg \u2022rali Fluency 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 \u2022upc-jmc \u2022 uedin-birch -0.5 -0.5 \u2022upv 23 24 25 26 27 28 29 30 31 32 19 20 21 22 23 24 25 26 27 28 117 In Domain Out of Domain Adequacy Adequacy German-English 15 16 17 18 19 20 21 22 23 24 25 26 27 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 lcc \u2022 upc-jmc \u2022systran \u2022upv Fluency \u2022ula \u2022upc-mr \u2022lcc 15 16 17 18 19 20 21 22 23 24 25 26 27 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 \u2022systran \u2022upv \u2022uedin-phi -jmc \u2022rali \u2022systran -0.3 -0.4 -0.5 -0.6 \u2022upv 12 13 14 15 16 17 18 19 20 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 Fluency uedin-phi \u2022 \u2022 \u2022utd \u2022upc-jmc \u2022upc-mr 0.4 \u2022rali -0.3 -0.4 -0.5 \u2022upv 12 13 14 15 16 17 18 19 20 0.3 0.2 0.1 -0.0 -0.1 -0.2 English-French In Domain Out of Domain Adequacy Adequacy .","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"We hypothesize that modeling morphological information will greatly constrain the set of possible tags, thereby further refining the representation of the tag lexicon.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"1.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"On the other hand, we are interested in the application of rhetorical analysis or \u00e2\u0080\u0098discourse parsing\u00e2\u0080\u0099 (3.2 and 3.3), in text generation (3.4), and in exploiting the corpus for the development of improved models of discourse structure (3.5).","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"Ex: Mr. Bush disclosed the policy by reading it...","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"Manual and Automatic Evaluation of Machine Translation between European Languages","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"Two general approaches are presented and two combination techniques are described for each approach.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"\u00c2\u00a3 : _ADV: 5.88 If:!","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"While there are other obstacles to completing this idea, we believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"It then computes a normalized Levenshtein edit distance between the extracted chain and the reference.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"8 1 8.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Since foreign names can be of any length, and since their original pronunciation is effectively unlimited, the identi\u00c2\u00ad fication of such names is tricky.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"Although these authors report better gains than ours, they are with respect to a non-adapted baseline.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"9 65.5 46.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"(2008) reported agreement between the teams (measured with Evalb) at 93.8% F1, the level of the CTB.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"This solution also obviates the need to perform word sense disambiguation.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"We consider the unsupervised POS induction problem without the use of a tagging dictionary.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"We model each parse as the decisions made to create it, and model those decisions as independent events.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Extract NE instance pairs with contexts First, we extract NE pair instances with their context from the corpus.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"The remaining arcs are marked OOV.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"The compressed variant uses block compression and is rather slow as a result.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"The results in Table 1 show PROBING is 81% faster than TRIE, which is in turn 31% faster than the fastest baseline.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Following this method, we repeatedly \u2014 say, 1000 times \u2014 sample sets of sentences from the output of each system, measure their BLEU score, and use these 1000 BLEU scores as basis for estimating a confidence interval.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Saving state allows our code to walk the data structure exactly once per query.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Unigram lookup is dense so we use an array of probability and backoff values.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"The test set included 2000 sentences from the Europarl corpus, but also 1064 sentences out-ofdomain test data.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"Models To assess the marginal utility of each component of the model (see Section 3), we incremen- tally increase its sophistication.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"Hyperparameter settings are sorted according to the median one-to-one metric over runs.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"Ideally we\u00e2\u0080\u0099d like to know the thematic role of each extracted noun phrase, but AutoSlog does not generate thematic roles.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"Systems that generally do better than others will receive a positive average normalizedjudgement per sentence.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Thus, rather than give a single evaluative score, we prefer to compare the performance of our method with the judgments of several human subjects.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"If \u00e2\u0080\u009cgun\u00e2\u0080\u009d and \u00e2\u0080\u009crevolver\u00e2\u0080\u009d refer to the same object, then it should also be acceptable to say that Fred was \u00e2\u0080\u009ckilled with a gun\u00e2\u0080\u009d and that the burglar \u00e2\u0080\u009cfireda revolver\u00e2\u0080\u009d.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"The accuracy of the sets in representing paraphrase ranged from 73% to 99%, depending on the NE categories and set sizes; the accuracy of the links for two evaluated domains was 73% and 86%.","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"The details of the transformation procedure are slightly different depending on the encoding schemes: d\u2191h let the linear head be the syntactic head). target arc must have the form wl \u2212\u2192 wm; if no target arc is found, Head is used as backoff. must have the form wl \u2212\u2192 wm and no outgoing arcs of the form wm p'\u2193 \u2212\u2192 wo; no backoff.","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"The 4th block contains instance-weighting models trained on all features, used within a MAP TM combination, and with a linear LM mixture.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"The complexity of the algorithm is O(E3 J2 2J), where E is the size of the target language vocabulary.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"There is twice as much language modelling data, since training data for the machine translation system is filtered against sentences of length larger than 40 words.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"In the named entity domain these rules were Each of these rules was given a strength of 0.9999.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"The zone to which a token belongs is used as a feature.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"This property reflects an important aspect of the underlying linguistic theory associated with the formalism.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"Gather phrases using keywords Next, we select a keyword for each phrase \u00e2\u0080\u0093 the top-ranked word based on the TF\/IDF metric.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"The logistic function, whose outputs are in [0, 1], forces pp(s, t) <_ po(s, t).","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"An important subproblem of language model storage is therefore sparse mapping: storing values for sparse keys using little memory then retrieving values given keys using little time.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"Our use of an unweighted lattice reflects our belief that all the segmentations of the given input sentence are a-priori equally likely; the only reason to prefer one segmentation over the another is due to the overall syntactic context which is modeled via the PCFG derivations.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"However, our full model takes advantage of word features not present in Grac\u00c2\u00b8a et al.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"G1 and G2 are hanzi, we can estimate the probability of the sequence being a name as the product of: \u00e2\u0080\u00a2 the probability that a word chosen randomly from a text will be a name-p(rule 1), and \u00e2\u0080\u00a2 the probability that the name is of the form 1hanzi-family 2hanzi-given-p(rule 2), and \u00e2\u0080\u00a2 the probability that the family name is the particular hanzi F1-p(rule 6), and \u00e2\u0080\u00a2 the probability that the given name consists of the particular hanzi G1 and G2-p(rule 9) This model is essentially the one proposed in Chang et al.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"In our model, however, all lattice paths are taken to be a-priori equally likely.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"Label propagation can then be used to transfer the labels to the peripheral foreign vertices (i.e. the ones adjacent to the English vertices) first, and then among all of the foreign vertices (\u00a74).","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"(a) ;IE shi4 'be' => ;IE;IE shi4bu2-shi4 (be-not-be) 'is it?'","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"Table 6: Example Translations for the Verbmobil task.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"This alters generation of T as follows: n P (T |\u00cf\u0088) = n P (Ti|\u00cf\u0088) i=1 Note that this distribution captures the frequency of a tag across word types, as opposed to tokens.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"In (2a), we want to split the two morphemes since the correct analysis is that we have the adverb :1 cai2 'just,' the modal verb neng2 'be able' and the main verb R: Hke4fu2 'overcome'; the competing analysis is, of course, that we have the noun :1 cai2neng2 'talent,' followed by }'lijke4fu2 'overcome.'","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"The second algorithm extends ideas from boosting algorithms, designed for supervised learning tasks, to the framework suggested by (Blum and Mitchell 98).","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"La ng ua ge 1T W + P RI O R + F E A T S E ng lis h D a ni s h D u tc h G e r m a n P or tu g u e s e S p a ni s h S w e di s h 2 1.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Table 9 shows that MADA produces a high quality segmentation, and that the effect of cascading segmentation errors on parsing is only 1.92% F1.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"A crucial difference is that the number of parameters is greatly reduced as is the number of variables that are sampled during each iteration.","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"In principle, it would be possible to encode the exact position of the syntactic head in the label of the arc from the linear head, but this would give a potentially infinite set of arc labels and would make the training of the parser very hard.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"Here, $ is the sentence boundary symbol, which is thought to be at position 0 in the target sentence.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"The learning task is to find two classifiers : 2x1 { \u20141, +1} 12 : 2x2 { \u20141, +1} such that (x1,) = f2(x2,t) = yt for examples i = 1, , m, and f1 (x1,) = f2 (x2,t) as often as possible on examples i = m + 1, ,n. To achieve this goal we extend the auxiliary function that bounds the training error (see Equ.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"In contrast, NNP (proper nouns) form a large portion of vocabulary.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"Information from a sentence is sometimes insufficient to classify a name correctly.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"In most cases, however, these expansions come with a steep increase in model complexity, with respect to training procedure and inference time.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"For example, in the phrase \u00e2\u0080\u009cCompany-A last week purchased rival Marshalls from Company-B\u00e2\u0080\u009d, the purchased company is Marshalls, not Company-B.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"We found that contextual role knowledge was more beneficial for pronouns than for definite noun phrases.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"We have developed a coreference resolver called BABAR that uses contextual role knowledge to make coreference decisions.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"This is done using a simple PCFG which is lexemebased.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"encodes the one tag per word constraint and is uni form over type-level tag assignments.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Generalizing state minimization, the model could also provide explicit bounds on probability for both backward and forward extension.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"One key issue here is to seek a discourse-based model of information structure.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"Similarly, the tokens and are tested against each list, and if found, a corresponding feature will be set to 1.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"One possibility is to use n-grams based on mutual information.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Finally, we note that simple weighting gives nearly a 2% F1 improvement, whereas Goldberg and Tsarfaty (2008) found that unweighted lattices were more effective for Hebrew.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Consider first the examples in (2).","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"0 57.2 43.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"While building a machine translation system is a serious undertaking, in future we hope to attract more newcomers to the field by keeping the barrier of entry as low as possible.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"First, we evaluated BABAR using only the seven general knowledge sources.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"The last affix in the list is the nominal plural f, men0.20 In the table are the (typical) classes of words to which the affix attaches, the number found in the test corpus by the method, the number correct (with a precision measure), and the number missed (with a recall measure).","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Replacing this with an ranked evaluation seems to be more suitable.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"The counts represent portions of the approximately 44000 constituents hypothesized by the parsers in the development set.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"For the PROBING implementation, hash table sizes are in the millions, so the most relevant values are on the right size of the graph, where linear probing wins.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"Each vertex within a connected component must have the same label \u2014 in the binary classification case, we need a single labeled example to identify which component should get which label.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"The German finite verbs 'bin' (second example) and 'konnten' (third example) are too far away from the personal pronouns 'ich' and 'Sie' (6 respectively 5 source sentence positions).","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"Our work is closest to that of Yarowsky and Ngai (2001), but differs in two important ways.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"We conjecture that this trend may continue by incorporating additional information, e.g., three-dimensional models as proposed by Tsarfaty and Sima\u2019an (2007).","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"The price shocked her.), or same-kind (e.g., Her health insurance paid for the hospital fees, but the automobile insurance did not cover the repair.).","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"Furthermore, the combination of pruning and vertical markovization of the grammar outperforms the Oracle results reported by Cohen and Smith.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"G1 and G2 are hanzi, we can estimate the probability of the sequence being a name as the product of: \u00e2\u0080\u00a2 the probability that a word chosen randomly from a text will be a name-p(rule 1), and \u00e2\u0080\u00a2 the probability that the name is of the form 1hanzi-family 2hanzi-given-p(rule 2), and \u00e2\u0080\u00a2 the probability that the family name is the particular hanzi F1-p(rule 6), and \u00e2\u0080\u00a2 the probability that the given name consists of the particular hanzi G1 and G2-p(rule 9) This model is essentially the one proposed in Chang et al.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"Two subjects are each given a calendar and they are asked to schedule a meeting.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"1 74.5 56.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"Despite their simplicity, uni- gram weights have been shown as an effective feature in segmentation models (Dyer, 2009).13 The joint parser\/segmenter is compared to a pipeline that uses MADA (v3.0), a state-of-the-art Arabic segmenter, configured to replicate ATB segmentation (Habash and Rambow, 2005).","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"The sentence length probability p(JjI) is omitted without any loss in performance.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"The kind of annotation work presented here would clearly benefit from the emergence of standard formats and tag sets, which could lead to sharable resources of larger size.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"Tsarfaty (2006) was the first to demonstrate that fully automatic Hebrew parsing is feasible using the newly available 5000 sentences treebank.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"In this case nonalpha is the string formed by removing all upper\/lower case letters from the spelling (e.g., for Thomas E. Petry nonalpha= .","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Finally, quite a few hanzi are homographs, meaning that they may be pronounced in several different ways, and in extreme cases apparently represent different morphemes: The prenominal modifi\u00c2\u00ad cation marker eg deO is presumably a different morpheme from the second morpheme of \u00c2\u00a7eg mu4di4, even though they are written the same way.4 The second point, which will be relevant in the discussion of personal names in Section 4.4, relates to the internal structure of hanzi.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"Restrictions: Quasi-monotone Search The above search space is still too large to allow the translation of a medium length input sentence.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Given the closeness of most systems and the wide over-lapping confidence intervals it is hard to make strong statements about the correlation between human judgements and automatic scoring methods such as BLEU.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"In addition to the optimizations specific to each datastructure described in Section 2, we implement several general optimizations for language modeling.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Different annotations of the same text are mapped into the same data structure, so that search queries can be formulated across annotation levels.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"17 They also provide a set of title-driven rules to identify names when they occur before titles such as $t. 1: xianlshengl 'Mr.' or i:l:itr!J tai2bei3 shi4zhang3 'Taipei Mayor.'","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Almost all annotators reported difficulties in maintaining a consistent standard for fluency and adequacy judgements, but nevertheless most did not explicitly move towards a ranking-based evaluation.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"Across eight European languages, our approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"The segmenter handles the grouping of hanzi into words and outputs word pronunciations, with default pronunciations for hanzi it cannot group; we focus here primarily on the system's ability to segment text appropriately (rather than on its pronunciation abilities).","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"In the namedentity problem each example is a (spelling,context) pair.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"07 95.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"\u00e2\u0080\u00a2 Some tools would allow for the desired annotation mode, but are so complicated (they can be used for many other purposes as well) that annotators take a long time getting used to them.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"Our first algorithm is similar to Yarowsky's, but with some important modifications motivated by (Blum and Mitchell 98).","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"We adopted the MUC6 guidelines for evaluating coreference relationships based on transitivity in anaphoric chains.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"The test set included 2000 sentences from the Europarl corpus, but also 1064 sentences out-ofdomain test data.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"The next two rows are results of oracle experiments.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"Each vertex within a connected component must have the same label \u2014 in the binary classification case, we need a single labeled example to identify which component should get which label.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"The difference between the featureless model (+PRIOR) and our full model (+FEATS) is 13.6% and 7.7% average error reduction on best and median settings respectively.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"These conaUses lossy compression. bThe 8-bit quantized variant returned incorrect probabilities as explained in Section 3.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"\u2022 We evaluated translation from English, in addition to into English.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"However, until such standards are universally adopted in evaluating Chinese segmenters, claims about performance in terms of simple measures like percent correct should be taken with a grain of salt; see, again, Wu and Fung (1994) for further arguments supporting this conclusion.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"This is a rather important source of errors in name identifi\u00c2\u00ad cation, and it is not really possible to objectively evaluate a name recognition system without considering the main lexicon with which it is used.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"This is most severe with RandLM in the multi-threaded case, where each thread keeps a separate cache, exceeding the original model size.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"We considered using the MUC6 and MUC7 data sets, but their training sets were far too small to learn reliable co-occurrence statistics for a large set of contextual role relationships.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"na me =>2 ha nzi fa mi ly 1 ha nzi gi ve n 6.1 ha nzi fa mi ly => ha nz ii 7.2 ha nzi fa mi ly => ha nzi i ha nz ij 8.1 ha nzi gi ve n => ha nz ii 9.2 ha nzi giv en => ha nzi i ha nz ij The difficulty is that given names can consist, in principle, of any hanzi or pair of hanzi, so the possible given names are limited only by the total number of hanzi, though some hanzi are certainly far more likely than others.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"As can be seen in Figure 3, the phrases in the \u00e2\u0080\u009cagree\u00e2\u0080\u009d set include completely different relationships, which are not paraphrases.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"We also performed experiments to evaluate the impact of each type of contextual role knowledge separately.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"Finally, our U (unparsed) measure is used to report the number of sentences to which our system could not propose a joint analysis.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"The manual scores are averages over the raw unnormalized scores.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"(1997; 1999) do have some statistics that show how IdentiFinder performs when the training data is reduced.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"Surprisingly, this effect is much less obvious for out-of-domain test data.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"In Table 5 we present results from small test cor\u00c2\u00ad pora for the productive affixes handled by the current version of the system; as with names, the segmentation of morphologically derived words is generally either right or wrong.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"Our first algorithm is similar to Yarowsky's, but with some important modifications motivated by (Blum and Mitchell 98).","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"Another attempt at using global information can be found in (Borthwick, 1999).","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"The domains are the linguistic spans that are to receive an IS-partitioning, and the units are the (smaller) spans that can play a role as a constituent of such a partitioning.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"The way we cant distinction between system performance.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"This is a simple and effective alternative to setting weights discriminatively to maximize a metric such as BLEU.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"As we shall see, most of the linked sets are paraphrases.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"If an unlabeled vertex does not have a path to any labeled vertex, this term ensures that the converged marginal for this vertex will be uniform over all tags, allowing the middle word of such an unlabeled vertex to take on any of the possible tags.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"\u00e2\u0080\u00a2 Bridging links: the annotator is asked to specify the type as part-whole, cause-effect (e.g., She had an accident.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"When keys are longer than 64 bits, we conserve space by replacing the keys with their 64-bit hashes.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"In recent years, coreference resolvers have been evaluated as part of MUC6 and MUC7 (MUC7 Proceedings, 1998).","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"Two common cases are the attribu tive adjective and the process nominal _; maSdar, which can have a verbal reading.4 At tributive adjectives are hard because they are or- thographically identical to nominals; they are inflected for gender, number, case, and definiteness.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Methods that allow multiple segmentations must provide criteria for choosing the best segmentation.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"5 64.7 42.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"The question of what soft function to pick, and how to design' algorithms which optimize it, is an open question, but appears to be a promising way of looking at the problem.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"4 69.0 51.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Although these are technically nominal, they have become known as \u00e2\u0080\u009cequational\u00e2\u0080\u009d sentences.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"We use MMAX for this annotation as well.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"This logic applies recursively: if wnf+1 similarly does not extend and has zero log backoff, it too should be omitted, terminating with a possibly empty context.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"For one thing, it is not clear who is to receive settlements or what should happen in case not enough teachers accept the offer of early retirement.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"To explore this tradeoff, we have performed experiments with three different encoding schemes (plus a baseline), which are described schematically in Table 1.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"The feature-HMM model works better for all languages, generalizing the results achieved for English by Berg-Kirkpatrick et al. (2010).","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"These are written to the state s(wn1) and returned so that they can be used for the following query.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"By establishing significantly higher parsing baselines, we have shown that Arabic parsing performance is not as poor as previously thought, but remains much lower than English.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"The test accuracy more or less asymptotes.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"One annotator suggested that this was the case for as much as 10% of our test sentences.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"In addition to the optimizations specific to each datastructure described in Section 2, we implement several general optimizations for language modeling.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"Evaluation metrics used are Attachment Score (AS), i.e. the proportion of tokens that are attached to the correct head, and Exact Match (EM), i.e. the proportion of sentences for which the dependency graph exactly matches the gold standard.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"annotation guidelines that tell annotators what to do in case of doubt.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"3 58.3 40.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"These sequence models-based approaches commonly treat token-level tag assignment as the primary latent variable.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"If the expression is longer or complicated (like \u00e2\u0080\u009cA buys B\u00e2\u0080\u009d and \u00e2\u0080\u009cA\u00e2\u0080\u0099s purchase of B\u00e2\u0080\u009d), it is called \u00e2\u0080\u009cparaphrase\u00e2\u0080\u009d, i.e. a set of phrases which express the same thing or event.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"We can better predict the probability of an unseen hanzi occurring in a name by computing a within-class Good-Turing estimate for each radical class.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"Also, the argument has been made that machine translation performance should be evaluated via task-based evaluation metrics, i.e. how much it assists performing a useful task, such as supporting human translators or aiding the analysis of texts.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Alon Lavie advised on this work.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"In recent years, coreference resolvers have been evaluated as part of MUC6 and MUC7 (MUC7 Proceedings, 1998).","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":".., Wn ) (obs) P T : Tag assigns (T1 ,.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"For each 13 Of course, this weighting makes the PCFG an improper distribution.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"We proposed an unsupervised method to discover paraphrases from a large untagged corpus.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Lexicon Feature: The string of the token is used as a feature.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"Furthermore, we do not connect the English vertices to each other, but only to foreign language vertices.4 The graph vertices are extracted from the different sides of a parallel corpus (De, Df) and an additional unlabeled monolingual foreign corpus Ff, which will be used later for training.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"More formally, we start by representing the dictionary D as a Weighted Finite State Trans\u00c2\u00ad ducer (WFST) (Pereira, Riley, and Sproat 1994).","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"If they knew that the first four words in a hypergraph node would never extend to the left and form a 5-gram, then three or even fewer words could be kept in the backward state.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"We develop our POS induction model based on the feature-based HMM of Berg-Kirkpatrick et al. (2010).","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"9 66.4 47.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"To this end, we picked 100 sentences at random containing 4,372 total hanzi from a test corpus.14 (There were 487 marks of punctuation in the test sentences, including the sentence-final periods, meaning that the average inter-punctuation distance was about 9 hanzi.)","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"(2009).","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"This enables us to build a high performance NER without using separate classifiers to take care of global consistency or complex formulation on smoothing and backoff models (Bikel et al., 1997).","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"t(i).","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"If is not initCaps, then (not-initCaps, ) is set to 1.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"input token, the segmentation is then performed deterministically given the 1-best analysis.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"Once the lexicon has been drawn, the model proceeds similarly to the standard token-level HMM: Emission parameters \u00ce\u00b8 are generated conditioned on tag assignments T . We also draw transition parameters \u00cf\u0086.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"A TAG consists of a finite set of elementary trees that are either initial trees or auxiliary trees.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Along with IRSTLM and TPT, our binary format is memory mapped, meaning the file and in-memory representation are the same.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"This suggests a strategy: run interpolation search until the range narrows to 4096 or fewer entries, then switch to binary search.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"Thus, the ATM has no more than 6km&quot; + 1 work tapes, where km&quot; is the maximum number of substrings spanned by a derived structure.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"The commentaries in PCC are all of roughly the same length, ranging from 8 to 10 sentences.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"First, the model assumes independence between the first and second hanzi of a double given name.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"gaolxing4 'happy'=> F.i'JF.i'JJI!JI!","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"For a language like English, this problem is generally regarded as trivial since words are delimited in English text by whitespace or marks of punctuation.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"The use of weighted transducers in particular has the attractive property that the model, as it stands, can be straightforwardly interfaced to other modules of a larger speech or natural language system: presumably one does not want to segment Chinese text for its own sake but instead with a larger purpose in mind.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"This section describes AdaBoost, which is the basis for the CoBoost algorithm.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Month Names, Days of the Week, and Numbers: If is initCaps and is one of January, February, . . .","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"Another approach to finding paraphrases is to find phrases which take similar subjects and objects in large corpora by using mutual information of word distribution [Lin and Pantel 01].","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"A less canonical representation of segmental morphology is triggered by a morpho-phonological process of omitting the definite article h when occurring after the particles b or l. This process triggers ambiguity as for the definiteness status of Nouns following these particles.We refer to such cases in which the concatenation of elements does not strictly correspond to the original surface form as super-segmental morphology.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"Unsupervised Part-of-Speech Tagging with Bilingual Graph-Based Projections","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"These tag distributions are used to initialize the label distributions over the English vertices in the graph.","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"However, when the linear LM is combined with a linear TM (lm+lin tm) or MAP TM (lm+map TM), the results are much better than a log-linear combination for the EMEA setting, and on a par for NIST.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Applications The discovered paraphrases have multiple applications.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"We then use linguistic and annotation insights to develop a manually annotated grammar for Arabic (\u00c2\u00a74).","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"The following auxiliary quantity is defined: Qe0 (e; C; j) := probability of the best partial hypothesis (ei 1; bi 1), where C = fbkjk = 1; ; ig, bi = j, ei = e and ei\u00f4\u0080\u0080\u00801 = e0.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"State will ultimately be used as context in a subsequent query.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"Dropping the conditioning on 0 for brevity, and letting \u00afc\u03bb(s, t) = c\u03bb(s, t) + yu(s|t), and \u00afc\u03bb(t) = 4Note that the probabilities in (7) need only be evaluated over the support of \u02dcp(s, t), which is quite small when this distribution is derived from a dev set.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Until now, all evaluations of Arabic parsing\u00e2\u0080\u0094including the experiments in the previous section\u00e2\u0080\u0094have assumed gold segmentation.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Local features are features that are based on neighboring tokens, as well as the token itself.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"BABAR merely identifies caseframes that frequently co-occur in coreference resolutions.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"2 56.2 32.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"The PROBING data structure is a rather straightforward application of these hash tables to store Ngram language models.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"More recently, Subramanya et al. (2010) defined a graph over the cliques in an underlying structured prediction model.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"It has been shown in (Chandra et al., 1981) that if M works in space log n then there is a deterministic TM which accepts the same language in polynomial time.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"However, the next step is clearly different.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"So, 1: f, xue2shengl+men0 (student+PL) 'students' occurs and we estimate its cost at 11.43; similarly we estimate the cost of f, jiang4+men0 (general+PL) 'generals' (as in 'J' f, xiao3jiang4+men0 'little generals'), at 15.02.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"The development of automatic scoring methods is an open field of research.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"At this step, we will try to link those sets, and put them into a single cluster.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Table 2 Similarity matrix for segmentation judgments.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Lexicalizing several POS tags improves performance.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"Table 5 shows the overall parsing accuracy attained with the three different encoding schemes, compared to the baseline (no special arc labels) and to training directly on non-projective dependency graphs.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"A corpus of German newspaper commentaries has been assembled and annotated with different information (and currently, to different degrees): part-of-speech, syntax, rhetorical structure, connectives, co-reference, and information structure.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"So, this was a surprise element due to practical reasons, not malice.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"We include a constituent in our hypothesized parse if it appears in the output of a majority of the parsers.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"Now we have sets of phrases which share a keyword and we have links between those sets.","label":0},{"source_text":"There is no global pruning.","target_text":"The approach recursively evaluates a quantity Q(C; j), where C is the set of already visited cities and sj is the last visited city.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"Xim, } associated with the ith example.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"Although these existential NPs do not need a prior referent, they may occur multiple times in a document.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"Still, for both human and automatic rhetorical analysis, connectives are the most important source of surface information.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"An important reason for separating the two types of features is that this opens up the possibility of theoretical analysis of the use of unlabeled examples.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"La ng ua ge # To ke ns # W or d Ty pe s # Ta gs E ng lis h D a ni s h D u tc h G e r m a n P or tu g u e s e S p a ni s h S w e di s h 1 1 7 3 7 6 6 9 4 3 8 6 2 0 3 5 6 8 6 9 9 6 0 5 2 0 6 6 7 8 8 9 3 3 4 1 9 1 4 6 7 4 9 2 0 6 1 8 3 5 6 2 8 3 9 3 7 2 3 2 5 2 8 9 3 1 1 6 4 5 8 2 0 0 5 7 4 5 2 5 1 2 5 4 2 2 4 7 4 1 Table 2: Statistics for various corpora utilized in experiments.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"We paid particular attention to minimize the number of free parameters, and used the same hyperparameters for all language pairs, rather than attempting language-specific tuning.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"Relative pronouns with only 1 NP in scope..","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"When finished, the whole material is written into an XML-structured annotation file.","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"Indeed there are several open issues.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"As in boosting, the algorithm works in rounds.","label":0},{"source_text":"Here we present two algorithms.","target_text":"For example, the independence assumptions mean that the model fails to capture the dependence between specific and more general features (for example the fact that the feature full.-string=New_York is always seen with the features contains (New) and The baseline method tags all entities as the most frequent class type (organization). contains (York) and is never seen with a feature such as contains (Group) ).","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"In this way we restrict the parameterization of a Language Original case English Danish Dutch German Spanish Swedish Portuguese 94.6 96.3 96.6 95.5 95.4 93.3 95.6 Table 1: Upper bound on tagging accuracy assuming each word type is assigned to majority POS tag.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"To evaluate proper-name identification, we randomly se\u00c2\u00ad lected 186 sentences containing 12,000 hanzi from our test corpus and segmented the text automatically, tagging personal names; note that for names, there is always a sin\u00c2\u00ad gle unambiguous answer, unlike the more general question of which segmentation is correct.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Preprocessing the raw trees improves parsing performance considerably.9 We first discard all trees dominated by X, which indicates errors and non-linguistic text.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"A derived structure will be mapped onto a sequence zi of substrings (not necessarily contiguous in the input), and the composition operations will be mapped onto functions that can defined as follows3. f((zi,\u2022 \u2022 \u2022 , zni), (m.,\u2022 \u2022 \u2022 ,Yn3)) = (Z1, \u2022 \u2022 \u2022 , Zn3) where each z, is the concatenation of strings from z,'s and yk's.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"(Hartmann 1984), for example, used the term Reliefgebung to characterize the distibution of main and minor information in texts (similar to the notion of nuclearity in RST).","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"In Section 4, we present the performance measures used and give translation results on the Verbmobil task.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"We developed a first version of annotation guidelines for co-reference in PCC (Gross 2003), which served as basis for annotating the core corpus but have not been empirically evaluated for inter-annotator agreement yet.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"The basic strategy is, for a given pair of entity types, to start with some examples, like several famous book title and author pairs; and find expressions which contains those names; then using the found expressions, find more author and book title pairs.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Linguistic intuitions like those in the previous section inform language-specific annotation choices.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"For the moment we will assume that there are only two possible labels: each y, is in { \u20141, +1}.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"We used the TF\/ITF metric to identify keywords.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"HR0011-06-C-0022.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"The idea is to have a pipeline of shallow-analysis modules (tagging, chunk- ing, discourse parsing based on connectives) and map the resulting underspecified rhetorical tree (see Section 2.4) into a knowledge base that may contain domain and world knowledge for enriching the representation, e.g., to resolve references that cannot be handled by shallow methods, or to hypothesize coherence relations.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"To check for this, we do pairwise bootstrap resampling: Again, we repeatedly sample sets of sentences, this time from both systems, and compare their BLEU scores on these sets.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"None of the models attach the attributive adjectives correctly.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Table 5: Individual Performance of KSs for Disasters (e.g., \u00e2\u0080\u009cthe mayor\u00e2\u0080\u009d vs. \u00e2\u0080\u009cthe journalist\u00e2\u0080\u009d).","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Some relations are signalled by subordinating conjunctions, which clearly demarcate the range of the text spans related (matrix clause, embedded clause).","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"We follow the guidelines developed in the TIGER project (Brants et al. 2002) for syntactic annotation of German newspaper text, using the Annotate3 tool for interactive construction of tree structures.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"For example, the phrase \u00e2\u0080\u009c's New York-based trust unit,\u00e2\u0080\u009d is not a paraphrase of the other phrases in the \u00e2\u0080\u009cunit\u00e2\u0080\u009d set.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"The system described in this paper is similar to the MENE system of (Borthwick, 1999).","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"We of course also fail to identify, by the methods just described, given names used without their associated family name.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Often, two systems can not be distinguished with a confidence of over 95%, so there are ranked the same.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Two of the Mainlanders also cluster close together but, interestingly, not particularly close to the Taiwan speakers; the third Mainlander is much more similar to the Taiwan speakers.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"This section describes AdaBoost, which is the basis for the CoBoost algorithm.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"The manual evaluation of scoring translation on a graded scale from 1\u20135 seems to be very hard to perform.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"A modified language model probability p\u00c3\u0086(eje0; e00) is defined as follows: p\u00c3\u0086(eje0; e00) =  1:0 if \u00c3\u0086 = 0 p(eje0; e00) if \u00c3\u0086 = 1 : We associate a distribution p(\u00c3\u0086) with the two cases \u00c3\u0086 = 0 and \u00c3\u0086 = 1 and set p(\u00c3\u0086 = 1) = 0:7.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Both of these analyses are shown in Figure 4; fortunately, the correct analysis is also the one with the lowest cost, so it is this analysis that is chosen.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"The results clearly indicate increased variation in the ATB relative to the WSJ, but care should be taken in assessing the magnitude of the difference.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"If one system is perfect, another has slight flaws and the third more flaws, a judge is inclined to hand out judgements of 5, 4, and 3.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"nan2gual 'pumpkin.'","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"including Third Tone Sandhi (Shih 1986), which changes a 3 (low) tone into a 2 (rising) tone before another 3 tone: 'j\";gil, xiao3 [lao3 shu3] 'little rat,' becomes xiao3 { lao2shu3 ], rather than xiao2 { lao2shu3 ], because the rule first applies within the word lao3shu3 'rat,' blocking its phrasal application.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"In total, for the 2,000 NE category pairs, 5,184 keywords are found.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"We have tested the translation system on the Verbmobil task (Wahlster 1993).","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"An interesting observation is that while vertical markovization benefits all our models, its effect is less evident in Cohen and Smith.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"Because the information flow in our graph is asymmetric (from English to the foreign language), we use different types of vertices for each language.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Another way to view the judgements is that they are less quality judgements of machine translation systems per se, but rankings of machine translation systems.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"Same for . In the case where the next token is a hyphen, then is also used as a feature: (init- Caps, ) is set to 1.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"Previous work has attempted to incorporate such constraints into token-level models via heavy-handed modifications to inference procedure and objective function (e.g., posterior regularization and ILP decoding) (Grac\u00c2\u00b8a et al., 2009; Ravi and Knight, 2009).","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"8 66.4 52.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"In Input: {(x1,i, Initialize: Vi, j : e(xi) = 0.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"This PCFG is incorporated into the Stanford Parser, a factored model that chooses a 1-best parse from the product of constituency and dependency parses.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"Having defined LCFRS's, in Section 4.2 we established the semilinearity (and hence constant growth property) of the languages generated.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"The use of global features has improved the performance on MUC6 test data from 90.75% to 93.27% (27% reduction in errors), and the performance on MUC7 test data from 85.22% to 87.24% (14% reduction in errors).","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"In a model we built with default settings, 1.2% of n + 1-grams were missing their ngram suffix.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"The first concerns how to deal with ambiguities in segmentation.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"then define the best segmentation to be the cheapest or best path in Id(I) o D* (i.e., Id(I) composed with the transitive closure of 0).6 Consider the abstract example illustrated in Figure 2.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"We include a constituent in our hypothesized parse if it appears in the output of a majority of the parsers.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"We compared the ATB5 to tree- banks for Chinese (CTB6), German (Negra), and English (WSJ) (Table 4).","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"The segmentation chosen is the best path through the WFST, shown in (d).","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"Among machine learning-based NERs, Identi- Finder has proven to be the best on the official MUC6 and MUC7 test data.","label":0},{"source_text":"There is no global pruning.","target_text":"Table 6: Example Translations for the Verbmobil task.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"A graph D = (W, A) is well-formed iff it is acyclic and connected.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"In this section we present a partial evaluation of the current system, in three parts.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"2 60.7 56.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Due to many similarly performing systems, we are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","label":1},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"The parser builds dependency graphs by traversing the input from left to right, using a stack to store tokens that are not yet complete with respect to their dependents.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"4.3 Translation Experiments.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"(2009) on Portuguese (Grac\u00c2\u00b8a et al.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"To see this, note thai the first two terms in the above equation correspond to the function that AdaBoost attempts to minimize in the standard supervised setting (Equ.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"The techniques we develop can be extended in a relatively straightforward manner to the more general case when OUT consists of multiple sub-domains.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"By taking the ratio of matching n-grams to the total number of n-grams in the system output, we obtain the precision pn for each n-gram order n. These values for n-gram precision are combined into a BLEU score: The formula for the BLEU metric also includes a brevity penalty for too short output, which is based on the total number of words in the system output c and in the reference r. BLEU is sensitive to tokenization.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"It differs from previous machine learning-based NERs in that it uses information from the whole document to classify each word, with just one classifier.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"We elected run Moses single-threaded to minimize the impact of RandLM\u2019s cache on memory use.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"the number of permutations carried out for the word reordering is given.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Jud ges A G G R ST M 1 M 2 M 3 T1 T2 T3 AG 0.7 0 0.7 0 0 . 4 3 0.4 2 0.6 0 0.6 0 0.6 2 0.5 9 GR 0.9 9 0 . 6 2 0.6 4 0.7 9 0.8 2 0.8 1 0.7 2 ST 0 . 6 4 0.6 7 0.8 0 0.8 4 0.8 2 0.7 4 M1 0.7 7 0.6 9 0.7 1 0.6 9 0.7 0 M2 0.7 2 0.7 3 0.7 1 0.7 0 M3 0.8 9 0.8 7 0.8 0 T1 0.8 8 0.8 2 T2 0.7 8 respectively, the recall and precision.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"The following auxiliary quantity is defined: Qe0 (e; C; j) := probability of the best partial hypothesis (ei 1; bi 1), where C = fbkjk = 1; ; ig, bi = j, ei = e and ei\u00f4\u0080\u0080\u00801 = e0.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"Learned Tag Prior (PRIOR) We next assume there exists a single prior distribution \u00cf\u0088 over tag assignments drawn from DIRICHLET(\u00ce\u00b2, K ).","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"For example, in Figure 3, we can see that the phrases in the \u00e2\u0080\u009cbuy\u00e2\u0080\u009d, \u00e2\u0080\u009cacquire\u00e2\u0080\u009d and \u00e2\u0080\u009cpurchase\u00e2\u0080\u009d sets are mostly paraphrases.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"Out-of-domain test data is from the Project Syndicate web site, a compendium of political commentary.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"Our work is closest to that of Yarowsky and Ngai (2001), but differs in two important ways.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"We also report word type level accuracy, the fraction of word types assigned their majority tag (where the mapping between model state and tag is determined by greedy one-to-one mapping discussed above).5 For each language, we aggregate results in the following way: First, for each hyperparameter setting, evaluate three variants: The first model (1TW) only 4 Typically, the performance stabilizes after only 10 itera-.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"M(wi) = Li).","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"With some minor API changes, namely returning the length of the n-gram matched, it could also be faster\u2014though this would be at the expense of an optimization we explain in Section 4.1.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"The idea of distinguishing between general and domain-specific examples is due to Daum\u00b4e and Marcu (2006), who used a maximum-entropy model with latent variables to capture the degree of specificity.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"As Figure 1 shows, this word has no high-confidence alignment in the Italian-English bitext.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"Two annotators received training with the RST definitions and started the process with a first set of 10 texts, the results of which were intensively discussed and revised.","label":1},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"Denote by g3(x) = Et crithl(x) , j E {1,2} the unthresholded strong-hypothesis (i.e., f3 (x) = sign(gi (x))).","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"For all these annotation taks, G\u00c2\u00a8otze developed a series of questions (essentially a decision tree) designed to lead the annotator to the ap propriate judgement.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"The percentage scores on the axis labels represent the amount of variation in the data explained by the dimension in question.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"The baseline system in Table 3 refers to the maximum entropy system that uses only local features.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"It has been shown for English (Wang and Hirschberg 1992; Hirschberg 1993; Sproat 1994, inter alia) that grammatical part of speech provides useful information for these tasks.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"5 \u00e2\u0080\u0098Underspecified Rhetorical Markup Language\u00e2\u0080\u0099 6 This confirms the figure given by (Schauer, Hahn.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"After tagging a large corpus with an automatic NE tagger, the method tries to find sets of paraphrases automatically without being given a seed phrase or any kinds of cue.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"This is because our corpus is not annotated, and hence does not distinguish between the various words represented by homographs, such as, which could be \/adv jiangl 'be about to' orInc jiang4 '(military) general'-as in 1j\\xiao3jiang4 'little general.'","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"(Berland and Charniak 99) describe a method for extracting parts of objects from wholes (e.g., &quot;speedometer&quot; from &quot;car&quot;) from a large corpus using hand-crafted patterns.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"We indicate whether a context with zero log backoff will extend using the sign bit: +0.0 for contexts that extend and \u22120.0 for contexts that do not extend.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"A better approach would be to distin guish between these cases, possibly by drawing on the vast linguistic work on Arabic connectives (AlBatal, 1990).","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"3 58.3 40.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"The authors acknowledge the support of the NSF (CAREER grant IIS0448168, and grant IIS 0904684).","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"4.2 A Sample Segmentation Using Only Dictionary Words Figure 4 shows two possible paths from the lattice of possible analyses of the input sentence B X:\u00c2\u00a5 .:.S:P:l 'How do you say octopus in Japanese?' previously shown in Figure 1.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"There has also been some work on adapting the word alignment model prior to phrase extraction (Civera and Juan, 2007; Wu et al., 2005), and on dynamically choosing a dev set (Xu et al., 2007).","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"All features were conjoined with the state z.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"For terrorism, BABAR generated 5,078 resolutions: 2,386 from lexical seeding and 2,692 from syntactic seeding.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Email: cls@bell-labs.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"Before explaining our method in detail, we present a brief overview in this subsection.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Not every annotator was fluent in both the source and the target language.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"The Wang, Li, and Chang system fails on fragment (b) because their system lacks the word youlyoul 'soberly' and misinterpreted the thus isolated first youl as being the final hanzi of the preceding name; similarly our system failed in fragment (h) since it is missing the abbreviation i:lJI!","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"This is a simple and effective alternative to setting weights discriminatively to maximize a metric such as BLEU.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"For example, out of 905 phrases in the CC- domain, 211 phrases contain keywords found in step 2.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"2 for the accuracy of the different methods.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"First, the model assumes independence between the first and second hanzi of a double given name.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"Le m\u00b4edicament de r\u00b4ef\u00b4erence de Silapo est EPREX\/ERYPO, qui contient de l\u2019\u00b4epo\u00b4etine alfa.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"Data We use the Hebrew Treebank, (Sima\u2019an et al., 2001), provided by the knowledge center for processing Hebrew, in which sentences from the daily newspaper \u201cHa\u2019aretz\u201d are morphologically segmented and syntactically annotated.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"The idea is to have a pipeline of shallow-analysis modules (tagging, chunk- ing, discourse parsing based on connectives) and map the resulting underspecified rhetorical tree (see Section 2.4) into a knowledge base that may contain domain and world knowledge for enriching the representation, e.g., to resolve references that cannot be handled by shallow methods, or to hypothesize coherence relations.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"In particular, the decision to represent arguments in verb- initial clauses as VP internal makes VSO and VOS configurations difficult to distinguish.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"We will refer to our system as MENERGI (Maximum Entropy Named Entity Recognizer using Global Information).","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"We will return to these issues in the discussion section.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"This is especially effective at reducing load time, since raw bytes are read directly to memory\u2014or, as happens with repeatedly used models, are already in the disk cache.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"There has been a lot of research on such lexical relations, along with the creation of resources such as WordNet.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"We will describe the evaluation of such clusters in the next subsection.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"During the search process, a partial hypothesis is extended by choosing a source sentence position, which has not been aligned with a target sentence position yet.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"Coverage indicates the fraction of hypotheses in which the character yield exactly matched the reference.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"computing the recall of the other's judgments relative to this standard.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"We retain segmentation markers\u00e2\u0080\u0094which are consistent only in the vocalized section of the treebank\u00e2\u0080\u0094to differentiate between e.g. \u00ef\u00bf\u00bd \u00e2\u0080\u009cthey\u00e2\u0080\u009d and \u00ef\u00bf\u00bd + \u00e2\u0080\u009ctheir.\u00e2\u0080\u009d Because we use the vocalized section, we must remove null pronoun markers.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"As noted in Section 4.4, disk cache state is controlled by reading the entire binary file before each test begins.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"To facilitate comparison with previous work, we exhaustively evaluate this grammar and two other parsing models when gold segmentation is assumed (\u00c2\u00a75).","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"While automatic measures are an invaluable tool for the day-to-day development of machine translation systems, they are only a imperfect substitute for human assessment of translation quality, or as the acronym BLEU puts it, a bilingual evaluation understudy.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"This variant is tested in the experiments below.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"(2010), we adopt a simpler na\u00c2\u00a8\u00c4\u00b1ve Bayes strategy, where all features are emitted independently.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"Syntactic universals are a well studied concept in linguistics (Carnie, 2002; Newmeyer, 2005), and were recently used in similar form by Naseem et al. (2010) for multilingual grammar induction.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"yields.6 But to its great advantage, it has a high ratio of non-terminals\/terminals (\u00ce\u00bc Constituents \/ \u00ce\u00bc Length).","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"However, when the linear LM is combined with a linear TM (lm+lin tm) or MAP TM (lm+map TM), the results are much better than a log-linear combination for the EMEA setting, and on a par for NIST.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"In order to evaluate and advance this approach, it helps to feed into the knowledge base data that is already enriched with some of the desired information \u00e2\u0080\u0094 as in PCC.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"This is summarized in Equation 5.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"TAG's can be used to give the structural descriptions discussed by Gazdar (1985) for the unbounded nested dependencies in Norwedish, for cross serial dependencies in Dutch subordinate clauses, and for the nestings of paired English complementizers.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"La ng ua ge 1T W + P RI O R + F E A T S E ng lis h D a ni s h D u tc h G e r m a n P or tu g u e s e S p a ni s h S w e di s h 2 1.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"But in most cases they can be used interchangably.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"BABAR merely identifies caseframes that frequently co-occur in coreference resolutions.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"2.2 Contextual Role Knowledge.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"It is difficult to motivate these days why one ministry should be exempt from cutbacks \u00e2\u0080\u0094 at the expense of the others.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"However, the next step is clearly different.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"In our grammar, features are realized as annotations to basic category labels.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"For instance: if 10 systems participate, and one system does better than 3 others, worse then 2, and is not significant different from the remaining 4, its rank is in the interval 3\u20137.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"An alternate approximation to (8) would be to let w,\\(s, t) directly approximate p\u02c6I(s, t).","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"If is a number string (such as one, two, etc), then the feature NumberString is set to 1.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"In an existential state an ATM behaves like a nondeterministic TM, accepting if one of the applicable moves leads to acceptance; in an universal state the ATM accepts if all the applicable moves lead to acceptance.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"(Bar-Haim et al., 2007; Habash and Rambow, 2005)) and probabilities are assigned to different analyses in accordance with the likelihood of their tags (e.g., \u201cfmnh is 30% likely to be tagged NN and 70% likely to be tagged REL+VB\u201d).","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"The NIST06 and NIST08 evaluation sets were used for testing.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"Whereas Arabic linguistic theory as Saddam (a) Reference (b) Stanford signs (1) and (2) to the class of pseudo verbs 01 +i J>1\u00ef\u00bf\u00bd inna and her sisters since they can beinflected, the ATB conventions treat (2) as a com plementizer, which means that it must be the head of SBAR.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"T e r r o r i s m Ca sef ra me Semantic Classes <a ge nt > ass ass ina ted group, human inv esti gat ion int o < N P> event exp lod ed out sid e < N P> building N a t u r a l D i s a s t e r s Ca sef ra me Semantic Classes <a ge nt > inv esti gat ing cau se group, human sur viv or of < N P> event, natphenom hit wit h < N P> attribute, natphenom Figure 3: Semantic Caseframe Expectations Figure 2: Lexical Caseframe Expectations To illustrate how lexical expectations are used, suppose we want to determine whether noun phrase X is the antecedent for noun phrase Y. If they are coreferent, then X and Y should be substitutable for one another in the story.4 Consider these sentences: (S1) Fred was killed by a masked man with a revolver.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"It was then tested on section 22 of the Treebank in conjunction with the other parsers.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"2 61.7 64.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"The task can be considered to be one component of the MUC (MUC-6, 1995) named entity task (the other task is that of segmentation, i.e., pulling possible people, places and locations from text before sending them to the classifier).","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"One conclusion drawn from this annotation effort was that for humans and machines alike, 2 www.sfs.nphil.unituebingen.de\/Elwis\/stts\/ stts.html 3 www.coli.unisb.de\/sfb378\/negra-corpus\/annotate.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"This is demonstrated by average scores over all systems, in terms of BLEU, fluency and adequacy, as displayed in Figure 5.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"This flexibility, along with the simplicity of implementation and expansion, makes this framework an attractive base for continued research.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"There are two possible reasons: (1) the knowledge sources are resolving different cases of anaphora, and (2) the knowledge sources provide multiple pieces of evidence in support of (or against) a candidate, thereby acting synergistically to push the DempsterShafer model over the belief threshold in favor of a single candidate.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"Because we are working with only three parsers, the only situation in which context will help us is when it can indicate we should choose to believe a single parser that disagrees with the majority hypothesis instead of the majority hypothesis itself.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"At first glance, we quickly recognize that many systems are scored very similar, both in terms of manual judgement and BLEU.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"Also, the method of using keywords rules out phrases which don\u00e2\u0080\u0099t contain popular words in the domain.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"For comparison, the completely unsupervised feature-HMM baseline accuracy on the universal POS tags for English is 79.4%, and goes up to 88.7% with a treebank dictionary.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":":zhong1 : 0.0 tjl :huo2 :0.0 (R:spub:\/ic of Ch:ina) + .,_,...I : jlong4 :0.0 (mUifaty genG181) 0 \u00c2\u00a3: _NC: 40.0 Figure 3 Partial Chinese Lexicon (NC = noun; NP = proper noun).c=- - I \u00e2\u0080\u00a2=- :il: .;ss:;zhangt \u00e2\u0080\u00a2 '-:.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"However, MADA is language-specific and relies on manually constructed dictionaries.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"The 95% confidence intervals for type-level errors are (5580, 9440) for the ATB and (1400, 4610) for the WSJ.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Part of the gap between resident and virtual memory is due to the time at which data was collected.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"The maximum likelihood estimates (i.e., parameter values which maximize 10) can not be found analytically, but the EM algorithm can be used to hill-climb to a local maximum of the likelihood function from some initial parameter settings.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"The first two rows of the table are baselines.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"i=1 (f,v)\u00e2\u0088\u0088Wi","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"These estimates are in turn combined linearly with relative-frequency estimates from an in-domain phrase table.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"We are currently exploring other methods that employ similar ideas and their formal properties.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"We observe similar trends when using another measure \u00e2\u0080\u0093 type-level accuracy (defined as the fraction of words correctly assigned their majority tag), according to which La ng ua ge M etr ic B K 10 E M B K 10 L B F G S G 10 F EA T S B es t F EA T S M ed ia n E ng lis h 1 1 m 1 4 8 . 3 6 8 . 1 5 6 . 0 7 5 . 5 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 0 . 9 6 6 . 4 4 7 . 8 6 6 . 4 D an is h 1 1 m 1 4 2 . 3 6 6 . 7 4 2 . 6 5 8 . 0 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 2 . 1 6 1 . 2 4 3 . 2 6 0 . 7 D ut ch 1 1 m 1 5 3 . 7 6 7 . 0 5 5 . 1 6 4 . 7 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 6 . 4 6 9 . 0 5 1 . 5 6 7 . 3 Po rtu gu es e 1 1 m 1 5 0 . 8 7 5 . 3 4 3 . 2 7 4 . 8 44 .5 69 .2 6 4 . 1 7 4 . 5 5 6 . 5 7 0 . 1 S pa ni sh 1 1 m 1 \u00e2\u0080\u0093 \u00e2\u0080\u0093 4 0 . 6 7 3 . 2 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 8 . 3 6 8 . 9 5 0 . 0 5 7 . 2 Table 4: Comparison of our method (FEATS) to state-of-the-art methods.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models (e.g., (Grosz et al., 1995; Grosz and Sidner, 1998)), syntactic algorithms (e.g., (Hobbs, 1978; Lappin and Le- ass, 1994)), and supervised machine learning systems (Aone and Bennett, 1995; McCarthy and Lehnert, 1995; Ng and Cardie, 2002; Soon et al., 2001).","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"The dictionary sizes reported in the literature range from 17,000 to 125,000 entries, and it seems reasonable to assume that the coverage of the base dictionary constitutes a major factor in the performance of the various approaches, possibly more important than the particular set of methods used in the segmentation.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"We feel that information from a whole corpus might turn out to be noisy if the documents in the corpus are not of the same genre.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"splitPUNC restores the convention of the WSJ.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"The results described in this paper suggest a characterization of mild context-sensitivity in terms of generalized context-freeness.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"Section 4 draws some conclusions from the present state of the effort.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"On the one hand, the type-level error rate is not calibrated for the number of n-grams in the sample.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"This is the parse that is closest to the centroid of the observed parses under the similarity metric.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"Its only purpose is 3 This follows since each \u00ce\u00b8t has St \u00e2\u0088\u0092 1 parameters and.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"We also collapse unary chains withidentical basic categories like NP \u00e2\u0086\u0092 NP.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"Search CPU time mWER SSER Method [sec] [%] [%] MonS 0:9 42:0 30:5 QmS 10:6 34:4 23:8 IbmS 28:6 38:2 26:2 4.2 Performance Measures.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"To explicitly handle the word reordering between words in source and target language, we use the concept of the so-called inverted alignments as given in (Ney et al., 2000).","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"2.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"(2009) study related but different multilingual grammar and tagger induction tasks, where it is assumed that no labeled data at all is available.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"To lower the barrier of entrance to the competition, we provided a complete baseline MT system, along with data resources.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"As far as we know, no other NERs have used information from the whole document (global) as well as information within the same sentence (local) in one framework.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"The first m pairs have labels yi, whereas for i = m + 1, , n the pairs are unlabeled.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"collected too.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"The frequency of the Company \u00e2\u0080\u0093 Company domain ranks 11th with 35,567 examples.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"1,000 of these were picked at random, and labeled by hand to produce a test set.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"For example, both a chair and a truck would be labeled as artifacts, but this does not at all suggest that they are coreferent.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"SRILM\u2019s compact variant has an incredibly expensive destructor, dwarfing the time it takes to perform translation, and so we also modified Moses to avoiding the destructor by calling exit instead of returning normally.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Next, we describe four contextual role knowledge sources that are created from the training examples and the caseframes.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Much confusion has been sown about Chinese writing by the use of the term ideograph, suggesting that hanzi somehow directly represent ideas.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"In line with perplexity results from Table 1, the PROBING model is the fastest followed by TRIE, and subsequently other packages.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"While our method also enforces a singe tag per word constraint, it leverages the transition distribution encoded in an HMM, thereby benefiting from a richer representation of context.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"They first collect the NE instance pairs and contexts, just like our method.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"8 66.4 52.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"The results in Table 1 show PROBING is 81% faster than TRIE, which is in turn 31% faster than the fastest baseline.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"Our primary goal is to exploit the resources that are most appropriate for the task at hand, and our secondary goal is to allow for comparison of our models\u2019 performance against previously reported results.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"For sorted lookup, we compare interpolation search, standard C++ binary search, and standard C++ set based on red-black trees.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"In addition, as the CRF and PCFG look at similar sorts of information from within two inherently different models, they are far from independent and optimizing their product is meaningless.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"There are 13 types of extensions needed to describe the verbgroup reordering.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"37.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"49 99.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","label":1},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"We will evaluate various specific aspects of the segmentation, as well as the overall segmentation per\u00c2\u00ad formance.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Tokens tagged as PUNC are not discarded unless they consist entirely of punctuation.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"(2) Description Source Location Names http:\/\/www.timeanddate.com http:\/\/www.cityguide.travel-guides.com http:\/\/www.worldtravelguide.net Corporate Names http:\/\/www.fmlx.com Person First Names http:\/\/www.census.gov\/genealogy\/names Person Last Names Table 2: Sources of Dictionaries The McCann family . . .","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"In order to pass these constraints onto the parser, the lexical rules in the grammar are of the form pi \u2014* (si, pi) Parameter Estimation The grammar probabilities are estimated from the corpus using simple relative frequency estimates.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"Second, we treat the projected labels as features in an unsupervised model (\u00a75), rather than using them directly for supervised training.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Mandarin exhibits several such processes, including A-not-A question formation, il\u00c2\u00ad lustrated in (3a), and adverbial reduplication, illustrated in (3b): 3.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"We use the universal POS tagset of Petrov et al. (2011) in our experiments.10 This set C consists of the following 12 coarse-grained tags: NOUN (nouns), VERB (verbs), ADJ (adjectives), ADV (adverbs), PRON (pronouns), DET (determiners), ADP (prepositions or postpositions), NUM (numerals), CONJ (conjunctions), PRT (particles), PUNC (punctuation marks) and X (a catch-all for other categories such as abbreviations or foreign words).","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"keywords Step 3 Sets of phrases based on keywords Step 4 Links between sets of phrases All the contexts collected for a given domain are gathered in a bag and the TF\/ITF scores are calculated for all the words except stopwords in the bag.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"Here we do not submit to this view.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"Pseudo-code describing the generalized boosting algorithm of Schapire and Singer is given in Figure 1.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"The first value reports use immediately after loading while the second reports the increase during scoring. dBerkeleyLM is written in Java which requires memory be specified in advance.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"In German, the verbgroup usually consists of a left and a right verbal brace, whereas in English the words of the verbgroup usually form a sequence of consecutive words.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"One of such approaches uses comparable documents, which are sets of documents whose content are found\/known to be almost the same, such as different newspaper stories about the same event [Shinyama and Sekine 03] or different translations of the same story [Barzilay 01].","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"We have tested the translation system on the Verbmobil task (Wahlster 1993).","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Time includes all queries but excludes random number generation and data structure population.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"A very small excerpt from an Italian-English graph is shown in Figure 1.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"They return a value in the range [0,1], where 0 indicates neutrality and 1 indicates the strongest belief that the candidate and anaphor are coreferent.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"11 www.ling.unipotsdam.de\/sfb\/projekt a3.php 12 This step was carried out in the course of the diploma thesis work of David Reitter (2003), which de serves special mention here.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"None of the models attach the attributive adjectives correctly.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"Experiments in two domains showed that the contextual role knowledge improved coreference performance, especially on pronouns.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"Note that in our model, conditioned on T , there is precisely one t which has nonzero probability for the token component, since for each word, exactly one \u00ce\u00b8t has support.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"What both of these approaches presume is that there is a sin\u00c2\u00ad gle correct segmentation for a sentence, against which an automatic algorithm can be compared.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"On MUC6 data, MENERGI also achieves performance comparable to IdentiFinder when trained on similar amount of training data.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"Part-of-speech (POS) tag distributions are known to exhibit sparsity \u00e2\u0080\u0094 a word is likely to take a single predominant tag in a corpus.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"On the one hand, the definition of composition in Steedman (1985), which technically permits composition of functions with unbounded number of arguments, generates tree sets with dependent paths such as those shown in Figure 6.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"each word in the lexicon whether or not each string is actually an instance of the word in question.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"Both of the switching techniques, as well as the parametric hybridization technique were also shown to be robust when a poor parser was introduced into the experiments.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"37.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"A name already mentioned previously in a document may appear in abbreviated form when it is mentioned again later.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"It was filtered to retain the top 30 translations for each source phrase using the TM part of the current log-linear model.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"5.1 Data Sets.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Virginia) and -sia are normally transliterated as fbSi!","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"For example, if CFLex determines that the log- likelihood statistic for the co-occurrence of a particular noun and caseframe corresponds to the 90% confidence level, then CFLex returns .90 as its belief that the anaphor and candidate are coreferent.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"(Webber et al., 2003)).","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"In the maximum entropy framework, there is no such constraint.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Furthermore, by inverting the transducer so that it maps from phonemic transcriptions to hanzi sequences, one can apply the segmenter to other problems, such as speech recognition (Pereira, Riley, and Sproat 1994).","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"Even if an example like this is not labeled, it can be interpreted as a &quot;hint&quot; that Mr and president imply the same category.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"The domains are the linguistic spans that are to receive an IS-partitioning, and the units are the (smaller) spans that can play a role as a constituent of such a partitioning.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"Each xt E 2x is the set of features constituting the ith example.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Although these are technically nominal, they have become known as \u00e2\u0080\u009cequational\u00e2\u0080\u009d sentences.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Similarly, hanzi sharing the GHOST radical _m tend to denote spirits and demons, such as _m gui3 'ghost' itself, II: mo2 'demon,' and yan3 'nightmare.'","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"This alters generation of T as follows: n P (T |\u00cf\u0088) = n P (Ti|\u00cf\u0088) i=1 Note that this distribution captures the frequency of a tag across word types, as opposed to tokens.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"So we decided to use semantic class information only to rule out candidates.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"These are written to the state s(wn1) and returned so that they can be used for the following query.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","label":1},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"(b) 89 :1 t& tal de cai2neng2 hen3 he DE talent very 'He has great talent' f.b ga ol hig h While the current algorithm correctly handles the (b) sentences, it fails to handle the (a) sentences, since it does not have enough information to know not to group the sequences.ma3lu4 and?]cai2neng2 respectively.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Gan's solution depends upon a fairly sophisticated language model that attempts to find valid syntactic, semantic, and lexical relations between objects of various linguistic types (hanzi, words, phrases).","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"For example: McCann initiated a new global system.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"When keys are longer than 64 bits, we conserve space by replacing the keys with their 64-bit hashes.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"There any many techniques for improving language model speed and reducing memory consumption.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"Since the work tapes store integers (which can be written in binary) that never exceed the size of the input, no configuration has space exceeding 0(log n).","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"We have presented a new method for non-projective dependency parsing, based on a combination of data-driven projective dependency parsing and graph transformation techniques.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"One hybridization strategy is to let the parsers vote on constituents' membership in the hypothesized set.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"As we will see from Table 3, not much improvement is derived from this feature.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"Confidence Interval: To estimate confidence intervals for the average mean scores for the systems, we use standard significance testing.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"TPT Germann et al. (2009) describe tries with better locality properties, but did not release code.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"Without HSPELL-pruning, our simpler grammars are somewhat lagging behind, but as the grammars improve the gap is bridged.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"Mai.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"Graph construction for structured prediction problems such as POS tagging is non-trivial: on the one hand, using individual words as the vertices throws away the context necessary for disambiguation; on the other hand, it is unclear how to define (sequence) similarity if the vertices correspond to entire sentences.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"In (1) the sequencema3lu4 cannot be resolved locally, but depends instead upon broader context; similarly in (2), the sequence :::tcai2neng2 cannot be resolved locally: 1.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"41.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"Parameter Component As in the standard Bayesian HMM (Goldwater and Griffiths, 2007), all distributions are independently drawn from symmetric Dirichlet distributions: 2 Note that t and w denote tag and word sequences respectively, rather than individual tokens or tags.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"Typically, judges initially spent about 3 minutes per sentence, but then accelerate with experience.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"1 61.7 37.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"BABAR uses a DempsterShafer decision model (Stefik, 1995) to combine the evidence provided by the knowledge sources.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"3.1 Lexicon Component.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Among these are words derived by various productive processes, including: 1.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"In the next section we present an alternative approach that builds two classifiers while attempting to satisfy the above constraints as much as possible.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"Ex: Mr. Cristiani, president of the country ...","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"If any of the tokens from to is in Person-Prefix- List, then another feature Person-Prefix is set to 1.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"Recent work has made significant progress on unsupervised POS tagging (Me\u00c2\u00b4rialdo, 1994; Smith and Eisner, 2005; Haghighi and Klein, 2006; Johnson,2007; Goldwater and Griffiths, 2007; Gao and John son, 2008; Ravi and Knight, 2009).","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"We claim that no particular morphological segmentation is a-priory more likely for surface forms before exploring the compositional nature of syntactic structures, including manifestations of various long-distance dependencies.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"We introduce several new ideas.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"He used an additional maximum entropy classifier that tries to correct mistakes by using reference resolution.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"As can be seen in Figure 3, the phrases in the \u00e2\u0080\u009cagree\u00e2\u0080\u009d set include completely different relationships, which are not paraphrases.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"We received submissions from 14 groups from 11 institutions, as listed in Figure 2.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Although these are technically nominal, they have become known as \u00e2\u0080\u009cequational\u00e2\u0080\u009d sentences.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"The first point we need to address is what type of linguistic object a hanzi repre\u00c2\u00ad sents.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"The Penn Arabic Treebank (ATB) syntactic guidelines (Maamouri et al., 2004) were purposefully borrowed without major modification from English (Marcus et al., 1993).","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"The ATB annotation guidelines specify that proper nouns should be specified with a flat NP (a).","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"For the \u00e2\u0080\u0098core\u00e2\u0080\u0099 portion of PCC, we found that on average, 35% of the coherence relations in our RST annotations are explicitly signalled by a lexical connective.6 When adding the fact that connectives are often ambiguous, one has to conclude that prospects for an automatic analysis of rhetorical structure using shallow methods (i.e., relying largely on connectives) are not bright \u00e2\u0080\u0094 but see Sections 3.2 and 3.3 below.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"Instead of the names of elementary trees of a TAG, the nodes are labeled by a sequence of names of trees in an elementary tree set.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Ideally we\u00e2\u0080\u0099d like to know the thematic role of each extracted noun phrase, but AutoSlog does not generate thematic roles.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"Following this method, we repeatedly \u2014 say, 1000 times \u2014 sample sets of sentences from the output of each system, measure their BLEU score, and use these 1000 BLEU scores as basis for estimating a confidence interval.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Finally, we note that simple weighting gives nearly a 2% F1 improvement, whereas Goldberg and Tsarfaty (2008) found that unweighted lattices were more effective for Hebrew.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"In this work, we take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Entries for 2 < n < N store a vocabulary identifier, probability, backoff, and an index into the n + 1-gram table.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Using less training data than other systems, our NER is able to perform as well as other state-of-the-art NERs.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"At the very least, we are creating a data resource (the manual annotations) that may the basis of future research in evaluation metrics.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"This suggests that by generalizing the notion of context-freeness in CFG's, we can define a class of grammatical formalisms that manipulate more complex structures.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Figure 3 shows a small fragment of the WFST encoding the dictionary, containing both entries forjust discussed, g:t\u00c2\u00a5 zhonglhua2 min2guo2 (China Republic) 'Republic of China,' and i\u00c2\u00a5inl.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Previous work deals with this problem by correcting inconsistencies between the named entity classes assigned to different occurrences of the same entity (Borthwick, 1999; Mikheev et al., 1998).","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"(3)) to be defined over unlabeled as well as labeled instances.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Whether a language even has orthographic words is largely dependent on the writing system used to represent the language (rather than the language itself); the notion \"orthographic word\" is not universal.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"We of course also fail to identify, by the methods just described, given names used without their associated family name.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"The simplest approach involves scoring the various analyses by costs based on word frequency, and picking the lowest cost path; variants of this approach have been described in Chang, Chen, and Chen (1991) and Chang and Chen (1993).","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"95 B a s e li n e ( S e lf t a g ) 70 a l l B i k e l ( v 1 . 2 ) B a s e l i n e ( P r e t a g ) 7 0 a l l G o l d P O S 70 0.7 70 0.801 278 0.7 52 0.794 278 0.7 71 0.804 295 0.7 52 0.796 295 0.7 75 0.808 309 77.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"Trees derived by IG's exhibit a property that is not exhibited by the trees sets derived by TAG's or CFG's.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"Given a sufficient number of randomly drawn unlabeled examples (i.e., edges), we will induce two completely connected components that together span the entire graph.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"For example, if FCC and Federal Communications Commission are both found in a document, then Federal has A begin set to 1, Communications has A continue set to 1, Commission has A end set to 1, and FCC has A unique set to 1.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"Lexicon Feature of Previous and Next Token: The string of the previous token and the next token is used with the initCaps information of . If has initCaps, then a feature (initCaps, ) is set to 1.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"After tagging a large corpus with an automatic NE tagger, the method tries to find sets of paraphrases automatically without being given a seed phrase or any kind of cue.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"(f1; ;mg ; l) 2 (f1; ;mg n fl; l1g ; l0) !","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"Here, \u00e2\u0080\u009cEG\u00e2\u0080\u009d represents \u00e2\u0080\u009cEastern Group Plc\u00e2\u0080\u009d.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Note that the backoff model assumes that there is a positive correlation between the frequency of a singular noun and its plural.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"3 68.9 50.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"Our derivation is slightly different from the one presented in (Schapire and Singer 98) as we restrict at to be positive.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"Xim, } associated with the ith example.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"We assumed that such a contrastive assessment would be beneficial for an evaluation that essentially pits different systems against each other.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"To summarize, we provided: The performance of the baseline system is similar to the best submissions in last year\u2019s shared task.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Thus, rather than give a single evaluative score, we prefer to compare the performance of our method with the judgments of several human subjects.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"Without HSPELL-pruning, our simpler grammars are somewhat lagging behind, but as the grammars improve the gap is bridged.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"See Figure 3 for a screenshot of the evaluation tool.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"For developers of Statistical Machine Translation (SMT) systems, an additional complication is the heterogeneous nature of SMT components (word-alignment model, language model, translation model, etc.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Following Sproat and Shih (1990), performance for Chinese segmentation systems is generally reported in terms of the dual measures of precision and recalP It is fairly standard to report precision and recall scores in the mid to high 90% range.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"The latter arcs correspond to OOV words in English.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"Ignoring the identity of the target language words e and e0, the possible partial hypothesis extensions due to the IBM restrictions are shown in Table 2.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Memory usage is the same as with binary search and lower than with set.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"For instance, the common \"suffixes,\" -nia (e.g.,.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"(Carlson, Marcu 2001) responded to this situation with relatively precise (and therefore long!)","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"This suggests a strategy: run interpolation search until the range narrows to 4096 or fewer entries, then switch to binary search.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"Better grammars are shown here to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"The major problem for our seg\u00c2\u00ad menter, as for all segmenters, remains the problem of unknown words (see Fung and Wu [1994]).","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"0.2 0.1 0.0 -0.1 25 26 27 28 29 30 31 32 -0.2 -0.3 \u2022systran \u2022 ntt 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 20 21 22 23 24 25 26 Fluency Fluency \u2022systran \u2022nrc rali 25 26 27 28 29 30 31 32 0.2 0.1 0.0 -0.1 -0.2 -0.3 cme p \ufffd 20 21 22 23 24 25 26 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 Figure 14: Correlation between manual and automatic scores for English-French 119 In Domain Out of Domain \u2022upv Adequacy -0.9 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 \u2022upv 23 24 25 26 27 28 29 30 31 32 \u2022upc-mr \u2022utd \u2022upc-jmc \u2022uedin-birch \u2022ntt \u2022rali \u2022uedin-birch 16 17 18 19 20 21 22 23 24 25 26 27 Adequacy \u2022upc-mr 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 -0.9 -1.0 -1.1 English-Spanish Fluency \u2022ntt \u2022nrc \u2022rali \u2022uedin-birch -0.2 -0.3 -0.5 \u2022upv 16 17 18 19 20 21 22 23 24 25 26 27 -0.4 nr \u2022 rali Fluency -0.4 \u2022upc-mr utd \u2022upc-jmc -0.5 -0.6 \u2022upv 23 24 25 26 27 28 29 30 31 32 0.2 0.1 -0.0 -0.1 -0.2 -0.3 0.3 0.2 0.1 -0.0 -0.1 -0.6 -0.7 Figure 15: Correlation between manual and automatic scores for English-Spanish 120 English-German In Domain Out of Domain Adequacy Adequacy 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 \u2022upv -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 -0.9 \u2022upv 0.5 0.4 \u2022systran \u2022upc-mr \u2022 \u2022rali 0.3 \u2022ntt 0.2 0.1 -0.0 -0.1 \u2022systran \u2022upc-mr -0.9 9 10 11 12 13 14 15 16 17 18 19 6 7 8 9 10 11 Fluency 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 \u2022upv -0.5 \u2022upv \u2022systran \u2022upc-mr \u2022 Fluency 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 \u2022systran \u2022ntt","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"The first value reports use immediately after loading while the second reports the increase during scoring. dBerkeleyLM is written in Java which requires memory be specified in advance.","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"On the other hand, we can expect Head+Path to be the most useful representation for reconstructing the underlying non-projective dependency graph.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"52 77.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"so that 'door' would be and in this case the hanzi 7C, does not represent a syllable.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"And if one is interested in TIS, one would probably consider the single orthographic word ACL to consist of three phonological words-lei s'i d\/-corresponding to the pronunciation of each of the letters in the acronym.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"We further thank Dr. J.-S.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"As was explained in the results section, \u00e2\u0080\u009cstrength\u00e2\u0080\u009d or \u00e2\u0080\u009cadd\u00e2\u0080\u009d are not desirable keywords in the CC-domain.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"Corpus Step 1 NE pair instances Step 2 Step 1.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"we perform five runs with different random initialization of sampling state.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"9 66.4 47.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"We use the HSPELL9 (Har\u2019el and Kenigsberg, 2004) wordlist as a lexeme-based lexicon for pruning segmentations involving invalid segments.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"We used section 23 as the development set for our combining techniques, and section 22 only for final testing.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"Even with less training data, MENERGI outperforms Borthwick' s MENE + reference resolution (Borthwick, 1999).","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"We first give a brief overview of boosting algorithms.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"This decreases the statistical significance of our results compared to those studies.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"from the subset of the United Informatics corpus not used in the training of the models.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"The problem with these styles of evaluation is that, as we shall demonstrate, even human judges do not agree perfectly on how to segment a given text.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"In addition to the automatic methods, AG, GR, and ST, just discussed, we also added to the plot the values for the current algorithm using only dictionary entries (i.e., no productively derived words or names).","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"This kind of dependency arises from the use of the composition operation to compose two arbitrarily large categories.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"2.2.2 The Caseframe Network The first type of contextual role knowledge that BABAR learns is the Caseframe Network (CFNet), which identifies caseframes that co-occur in anaphor\/antecedent resolutions.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"The PROBING model is 2.4 times as fast as the fastest alternative, SRILM, and uses less memory too.","label":0},{"source_text":"There is no global pruning.","target_text":"Only one of the first n positions which are not already aligned in a partial hypothesis may be chosen, where n is set to 4.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"If somewhere else in the document we see \u00e2\u0080\u009crestrictions put in place by President Bush\u00e2\u0080\u009d, then we can be surer that Bush is a name.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"We model po(s|t) using a MAP criterion over weighted phrase-pair counts: and from the similarity to (5), assuming y = 0, we see that w\u03bb(s, t) can be interpreted as approximating pf(s, t)\/po(s, t).","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"In this way we restrict the parameterization of a Language Original case English Danish Dutch German Spanish Swedish Portuguese 94.6 96.3 96.6 95.5 95.4 93.3 95.6 Table 1: Upper bound on tagging accuracy assuming each word type is assigned to majority POS tag.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Out-of-Vocabulary: We derived a lexicon list from WordNet 1.6, and words that are not found in this list have a feature out-of-vocabulary set to 1.","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"Compared to related work on the recovery of long-distance dependencies in constituency-based parsing, our approach is similar to that of Dienes and Dubey (2003) in that the processing of non-local dependencies is partly integrated in the parsing process, via an extension of the set of syntactic categories, whereas most other approaches rely on postprocessing only.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"The feature-HMM model works better for all languages, generalizing the results achieved for English by Berg-Kirkpatrick et al. (2010).","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Recall.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"Lack of correct reference translations was pointed out as a short-coming of our evaluation.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Less frequently studied is the interplay among language, annotation choices, and parsing model design (Levy and Manning, 2003; Ku\u00c2\u00a8 bler, 2005).","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"After tagging a large corpus with an automatic NE tagger, the method tries to find sets of paraphrases automatically without being given a seed phrase or any kinds of cue.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"3.4 Salience-based text generation.","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"It is not immediately obvious how to formulate an equivalent to equation (1) for an adapted TM, because there is no well-defined objective for learning TMs from parallel corpora.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"For all languages we do not make use of a tagging dictionary.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Hence, there is a total of 29 classes (7 name classes 4 sub-classes 1 not-a-name class).","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"and \u00e2\u0080\u009cH\u00e2\u0080\u009d represents \u00e2\u0080\u009cHanson Plc\u00e2\u0080\u009d.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"The BLEU metric, as all currently proposed automatic metrics, is occasionally suspected to be biased towards statistical systems, especially the phrase-based systems currently in use.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"2.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"This scenario is applicable to a large set of languages and has been considered by a number of authors in the past (Alshawi et al., 2000; Xi and Hwa, 2005; Ganchev et al., 2009).","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"To optimize left-to-right queries, we extend state to store backoff information: where m is the minimal context from Section 4.1 and b is the backoff penalty.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"Step 3.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"This may be the sign of a maturing research environment.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"Obviously \u00e2\u0080\u009cLotus\u00e2\u0080\u009d is part of the following clause rather than being the object of \u00e2\u0080\u009cestimates\u00e2\u0080\u009d and the extracted instance makes no sense.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"Ex: The regime gives itself the right...","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"Future extensions of the system might include: 1) An extended translation model, where we use more context to predict a source word.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"The hypothesized parse is then the set of constituents that are likely (P > 0.5) to be in the parse according to this model.","label":0},{"source_text":"There is no global pruning.","target_text":"The inverted alignment probability p(bijbi\u00f4\u0080\u0080\u00801; I; J) and the lexicon probability p(fbi jei) are obtained by relative frequency estimates from the Viterbi alignment path after the final training iteration.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"In this case nonalpha is the string formed by removing all upper\/lower case letters from the spelling (e.g., for Thomas E. Petry nonalpha= .","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"We describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"For all grammars, we use fine-grained PoS tags indicating various morphological features annotated therein.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"7 68.3 56.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"We assumed that such a contrastive assessment would be beneficial for an evaluation that essentially pits different systems against each other.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"BABAR merely identifies caseframes that frequently co-occur in coreference resolutions.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"We are therefore applying a different method, which has been used at the 2005 DARPA\/NIST evaluation.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"However, there are several reasons why this approach will not in general work: 1.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"nan2gual 'pumpkin.'","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"In this case, we have no finite-state restrictions for the search space.","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"The motivation for our more informal approach was the intuition that there are so many open problems in rhetorical analysis (and more so for German than for English; see below) that the main task is qualitative investigation, whereas rigorous quantitative analyses should be performed at a later stage.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"(Specifically, the limit n starts at 5 and increases by 5 at each iteration.)","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"However, using the top-level semantic classes of WordNet proved to be problematic because the class distinctions are too coarse.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Statistics are printed before Moses exits and after parts of the decoder have been destroyed.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"For statistics on this test set, refer to Figure 1.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"IRST is not threadsafe.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Using RandLM and the documented settings (8-bit values and 1 256 false-positive probability), we built a stupid backoff model on the same data as in Section 5.2.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"This is by now a fairly standard representation for multiple morphological segmentation of Hebrew utterances (Adler, 2001; Bar-Haim et al., 2005; Smith et al., 2005; Cohen and Smith, 2007; Adler, 2007).","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":".., Wn ) (obs) P T : Tag assigns (T1 ,.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Not every annotator was fluent in both the source and the target language.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"We show how a datadriven deterministic dependency parser, in itself restricted to projective structures, can be combined with graph transformation techniques to produce non-projective structures.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"The average fluency judgement per judge ranged from 2.33 to 3.67, the average adequacy judgement ranged from 2.56 to 4.13.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"The local features used are similar to those used in BBN' s IdentiFinder (Bikel et al., 1999) or MENE (Borthwick, 1999).","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"If 7' is an initial tree we do not include an address on the left-hand side.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Several systems propose statistical methods for handling unknown words (Chang et al. 1992; Lin, Chiang, and Su 1993; Peng and Chang 1993).","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"First, in section 4, we evaluate the graph transformation techniques in themselves, with data from the Prague Dependency Treebank and the Danish Dependency Treebank.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"The normalized judgement per sentence is the raw judgement plus (0 minus average raw judgement for this judge on this sentence).","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"2 for the accuracy of the different methods.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"For example, a person\u00e2\u0080\u0099s full name will match with just their last name (e.g., \u00e2\u0080\u009cGeorge Bush\u00e2\u0080\u009d and \u00e2\u0080\u009cBush\u00e2\u0080\u009d), and a company name will match with and without a corporate suffix (e.g., \u00e2\u0080\u009cIBM Corp.\u00e2\u0080\u009d and \u00e2\u0080\u009cIBM\u00e2\u0080\u009d).","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"The algorithm builds two classifiers in parallel from labeled and unlabeled data.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Figure 5 shows how this model is implemented as part of the dictionary WFST.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"36.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"To this end, we construct a bilingual graph over word types to establish a connection between the two languages (\u00a73), and then use graph label propagation to project syntactic information from English to the foreign language (\u00a74).","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"The negative logarithm of t0 is reported.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"(b) does the translation have the same meaning, including connotations?","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"besuchen 9.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Any NLP application that presumes as input unrestricted text requires an initial phase of text analysis; such applications involve problems as diverse as machine translation, information retrieval, and text-to-speech synthesis (TIS).","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Benchmarks use the package\u2019s binary format; our code is also the fastest at building a binary file.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Almost all annotators expressed their preference to move to a ranking-based evaluation in the future.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"Figure 4 shows some such phrase sets based on keywords in the CC-domain.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Once we figure out the important word (e.g. keyword), we believe we can capture the meaning of the phrase by the keyword.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"And if one is interested in TIS, one would probably consider the single orthographic word ACL to consist of three phonological words-lei s'i d\/-corresponding to the pronunciation of each of the letters in the acronym.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"We extend Subramanya et al.\u2019s intuitions to our bilingual setup.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"Skipped (K): The translation of up to one word may be postponed . Verb (V): The translation of up to two words may be anticipated.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"It is natural that the larger the data in the domain, the more keywords are found.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"The number of judgements is additionally fragmented by our breakup of sentences into in-domain and out-of-domain.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Thus we have some confidence that our own performance is at least as good as that of Chang et al.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"The PROBING data structure is a rather straightforward application of these hash tables to store Ngram language models.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"This assumption, however, is not inherent to type-based tagging models.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"We have described grammar state splits that significantly improve parsing performance, catalogued parsing errors, and quantified the effect of segmentation errors.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"Using less training data than other systems, our NER is able to perform as well as other state-of-the-art NERs.","label":1},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"(1997; 1999) do have some statistics that show how IdentiFinder performs when the training data is reduced.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"The parser switching oracle is the upper bound on the accuracy that can be achieved on this set in the parser switching framework.","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"In the rhetorical tree, nuclearity information is then used to extract a \u00e2\u0080\u009ckernel tree\u00e2\u0080\u009d that supposedly represents the key information from which the summary can be generated (which in turn may involve co-reference information, as we want to avoid dangling pronouns in a summary).","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"Recent research has demonstrated that incorporating this sparsity constraint improves tagging accuracy.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"This is to allow for fair comparison between the statistical method and GR, which is also purely dictionary-based.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"Secondly, some segments in a proposed segment sequence may in fact be seen lexical events, i.e., for some p tag Prf(p \u2014* (s, p)) > 0, while other segments have never been observed as a lexical event before.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"10 Other orthographic normalization schemes have been suggested for Arabic (Habash and Sadat, 2006), but we observe negligible parsing performance differences between these and the simple scheme used in this evaluation.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"97 81.","label":0},{"source_text":"Here we present two algorithms.","target_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"We can model this probability straightforwardly enough with a probabilistic version of the grammar just given, which would assign probabilities to the individual rules.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"For each extension a new position is added to the coverage set.","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"A corpus of German newspaper commentaries has been assembled at Potsdam University, and annotated with different linguistic information, to different degrees.","label":1},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"The entries in this table can be compared with those of Table 3 to see how the performance of the combining techniques degrades in the presence of an inferior parser.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"(2010), we adopt a simpler na\u00c2\u00a8\u00c4\u00b1ve Bayes strategy, where all features are emitted independently.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"Moving beyond directly related work, major themes in SMT adaptation include the IR (Hildebrand et al., 2005; L\u00a8u et al., 2007; Zhao et al., 2004) and mixture (Finch and Sumita, 2008; Foster and Kuhn, 2007; Koehn and Schroeder, 2007; L\u00a8u et al., 2007) approaches for LMs and TMs described above, as well as methods for exploiting monolingual in-domain text, typically by translating it automatically and then performing self training (Bertoldi and Federico, 2009; Ueffing et al., 2007; Schwenk and Senellart, 2009).","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"0 57.3 51.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"The weak learner for two-class problems computes a weak hypothesis h from the input space into the reals (h : 2x -4 R), where the sign4 of h(x) is interpreted as the predicted label and the magnitude I h(x)I is the confidence in the prediction: large numbers for I h(x)I indicate high confidence in the prediction, and numbers close to zero indicate low confidence.","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"Compared to related work on the recovery of long-distance dependencies in constituency-based parsing, our approach is similar to that of Dienes and Dubey (2003) in that the processing of non-local dependencies is partly integrated in the parsing process, via an extension of the set of syntactic categories, whereas most other approaches rely on postprocessing only.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Hieu Hoang named the code \u201cKenLM\u201d and assisted with Moses along with Barry Haddow.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"The development of automatic scoring methods is an open field of research.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"There are still some open issues to be resolved with the format, but it represents a first step.","label":0},{"source_text":"Here we present two algorithms.","target_text":"Taking \/V, to be the number of examples an algorithm classified correctly (where all gold standard items labeled noise were counted as being incorrect), we calculated two measures of accuracy: See Tab.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Non-empty buckets contain an entry belonging to them or to a preceding bucket where a conflict occurred.","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"We call this pseudoprojective dependency parsing, since it is based on a notion of pseudo-projectivity (Kahane et al., 1998).","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"This fact annoyed especially his dog...).","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"This PP modifies another NP, whose head is a singular noun.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"Log-linear combination (loglin) improves on this in all cases, and also beats the pure IN system.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"37 84.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"We have estimated the performance of IdentiFinder ' 99 at 200K words of training data from the graphs.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"As noted, this sentence consists of four words, namely B X ri4wen2 'Japanese,' :\u00c2\u00a5, zhanglyu2 'octopus\/ :&P:l zen3me0 'how,' and IDt shuol 'say.'","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Collisions between two keys in the table can be identified at model building time.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"For this experiment, we compare our model with the uniform tag assignment prior (1TW) with the learned prior (+PRIOR).","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Finally, we model the probability of a new transliterated name as the product of PTN and PTN(hanzi;) for each hanzi; in the putative name.13 The foreign name model is implemented as an WFST, which is then summed with the WFST implementing the dictionary, morpho 13 The current model is too simplistic in several respects.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Along with IRSTLM and TPT, our binary format is memory mapped, meaning the file and in-memory representation are the same.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"We are interested in combining the substructures of the input parses to produce a better parse.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"The high 1 tone of J1l would not normally neutralize in this fashion if it were functioning as a word on its own.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"Manual and Automatic Evaluation of Machine Translation between European Languages","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"ni2ya3 and @5:2 xilya3, respectively.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"In each figure the upper graph shows the isolated constituent precision and the bottom graph shows the corresponding number of hypothesized constituents.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Minimal perfect hashing is used to find the index at which a quantized probability and possibly backoff are stored.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"The 13 positions of the source sentence are processed in the order shown.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"A few pointed out that adequacy should be broken up into two criteria: (a) are all source words covered?","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"The tool we use is MMAX8, which has been specifically designed for marking co-reference.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"For instance, for TTS it is necessary to know that a particular sequence of hanzi is of a particular category because that knowl\u00c2\u00ad edge could affect the pronunciation; consider, for example the issues surrounding the pronunciation of ganl I qian2 discussed in Section 1.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Proper-Name Identification.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"So, it is too costly to make IE technology \u00e2\u0080\u009copen- domain\u00e2\u0080\u009d or \u00e2\u0080\u009con-demand\u00e2\u0080\u009d like IR or QA.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"42 nator, the N31s can be measured well by counting, and we replace the expectation by the observation.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"The index in this array is the vocabulary identifier.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Caching for IRSTLM is smaller at 0.09 GB resident memory, though it supports only a single thread.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"Pseudo-labels are formed by taking seed labels on the labeled examples, and the output of the fixed classifier on the unlabeled examples.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"Having found (spelling, context) pairs in the parsed data, a number of features are extracted.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"The inverted alignment probability p(bijbi\u00f4\u0080\u0080\u00801; I; J) and the lexicon probability p(fbi jei) are obtained by relative frequency estimates from the Viterbi alignment path after the final training iteration.","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"Thus, the arc from je to jedna will be labeled 5b\u2193 (to indicate that there is a syntactic head below it).","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"(Brandt 1996) extended these ideas toward a conception of kommunikative Gewichtung (\u00e2\u0080\u0098communicative-weight assignment\u00e2\u0080\u0099).","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"For English POS tagging, BergKirkpatrick et al. (2010) found that this direct gradient method performed better (>7% absolute accuracy) than using a feature-enhanced modification of the Expectation-Maximization (EM) algorithm (Dempster et al., 1977).8 Moreover, this route of optimization outperformed a vanilla HMM trained with EM by 12%.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"suffixes (e.g., \u00ef\u00bf\u00bd=) Other notable parameters are second order vertical Markovization and marking of unary rules.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"When a comparison against previous results requires additional pre-processing, we state it explicitly to allow for the reader to replicate the reported results.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Similarly, hanzi sharing the GHOST radical _m tend to denote spirits and demons, such as _m gui3 'ghost' itself, II: mo2 'demon,' and yan3 'nightmare.'","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"newspaper material, but also including kungfu fiction, Buddhist tracts, and scientific material.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":", December, then the feature MonthName is set to 1.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Keys to the table are hashed, using for example Austin Appleby\u2019s MurmurHash2, to integers evenly distributed over a large range.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"Conversely, the lattice parser requires no linguistic resources and produces segmentations of comparable quality.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"We then show that the combining techniques presented above give better parsing accuracy than any of the individual parsers.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"In addition to the automatic methods, AG, GR, and ST, just discussed, we also added to the plot the values for the current algorithm using only dictionary entries (i.e., no productively derived words or names).","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"In (2a), we want to split the two morphemes since the correct analysis is that we have the adverb :1 cai2 'just,' the modal verb neng2 'be able' and the main verb R: Hke4fu2 'overcome'; the competing analysis is, of course, that we have the noun :1 cai2neng2 'talent,' followed by }'lijke4fu2 'overcome.'","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"Formally, the DempsterShafer theory defines a probability density function m(S), where S is a set of hypotheses.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Judges varied in the average score they handed out.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"The number of NE instance pairs used in their experiment is less than half of our method.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"We also removed the context-type feature type when using the CoBoost approach.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"We carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","label":1},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"There has also been some work on adapting the word alignment model prior to phrase extraction (Civera and Juan, 2007; Wu et al., 2005), and on dynamically choosing a dev set (Xu et al., 2007).","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"It should be clear from the onset that the particle b (\u201cin\u201d) in \u2018bcl\u2019 may then attach higher than the bare noun cl (\u201cshadow\u201d).","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"For a corpus C, let M be the set of tuples \u00e2\u0088\u0097n, l), where n is an n-gram with bracketing label l. If any n appears 6 Generative parsing performance is known to deteriorate with sentence length.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"Note that Zt is a normalization constant that ensures the distribution Dt+i sums to 1; it is a function of the weak hypothesis ht and the weight for that hypothesis at chosen at the tth round.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Each model was able to produce hypotheses for all input sentences.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"The question is how to normalize the probabilities in such a way that smaller groupings have a better shot at winning.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"(1991}, Gu and Mao (1994), and Nie, Jin, and Hannan (1994).","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"Lexicon Feature of Previous and Next Token: The string of the previous token and the next token is used with the initCaps information of . If has initCaps, then a feature (initCaps, ) is set to 1.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"2.2 Syntactic structure.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"Figure 4 shows some such phrase sets based on keywords in the CC-domain.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"For example, as Gan (1994) has noted, one can construct examples where the segmen\u00c2\u00ad tation is locally ambiguous but can be determined on the basis of sentential or even discourse context.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"However, lexically similar NPs usually refer to the same entity in two cases: proper names and existential noun phrases.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"Several coreference resolvers have used supervised learning techniques, such as decision trees and rule learners (Aone and Bennett, 1995; McCarthy and Lehnert, 1995; Ng and Cardie, 2002; Soon et al., 2001).","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Therefore in cases where the segmentation is identical between the two systems we assume that tagging is also identical.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"An inverted alignment is defined as follows: inverted alignment: i ! j = bi: Target positions i are mapped to source positions bi.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"The basic word order is VSO, but SVO, VOS, and VO configurations are also possible.2 Nouns and verbs are created by selecting a consonantal root (usually triliteral or quadriliteral), which bears the semantic core, and adding affixes and diacritics.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"Often, two systems can not be distinguished with a confidence of over 95%, so there are ranked the same.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"The overall performance of our joint framework demonstrates that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperforms upper bounds proposed by previous joint disambiguation systems and achieves segmentation and parsing results on a par with state-of-the-art standalone applications results.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"The ATB segmentation scheme is one of many alternatives.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Almost all annotators reported difficulties in maintaining a consistent standard for fluency and adequacy judgements, but nevertheless most did not explicitly move towards a ranking-based evaluation.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"More complex approaches such as the relaxation technique have been applied to this problem Fan and Tsai (1988}.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"One possibility is to use n-grams based on mutual information.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Only tokens with initCaps not found in commonWords are tested against each list in Table 2.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"yields.6 But to its great advantage, it has a high ratio of non-terminals\/terminals (\u00ce\u00bc Constituents \/ \u00ce\u00bc Length).","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"This problem arises because our keywords consist of only one word.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"The judgements tend to be done more in form of a ranking of the different systems.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"In fact, it is very difficult to maintain consistent standards, on what (say) an adequacy judgement of 3 means even for a specific language pair.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"We have shown the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Microsoft\u2019s approach uses dependency trees, others use hierarchical phrase models.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"This is not an oversight.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"A procedural definition to restrict1In the approach described in (Berger et al., 1996), a mor phological analysis is carried out and word morphemes rather than full-form words are used during the search.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"For a sentence x and a state sequence z, a first order Markov model defines a distribution: (9) where Val(X) corresponds to the entire vocabulary.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Our annotators pointed out that very often they made almost random decisions as to what relation to choose, and where to locate the boundary of a span.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"Furthermore, the systematic way in which particles are prefixed to one another and onto an open-class category gives rise to a distinct sort of morphological ambiguity: space-delimited tokens may be ambiguous between several different segmentation possibilities.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"All commentaries have been tagged with part-of-speech information using Brants\u00e2\u0080\u0099 TnT1 tagger and the Stuttgart\/Tu\u00c2\u00a8bingen Tag Set automatic analysis was responsible for this decision.)","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Chunking is not enough to find such relationships.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Space- or punctuation-delimited * 700 Mountain Avenue, 2d451, Murray Hill, NJ 07974, USA.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Turning now to (1), we have the similar problem that splitting.into.ma3 'horse' andlu4 'way' is more costly than retaining this as one word .ma3lu4 'road.'","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"If is a number string (such as one, two, etc), then the feature NumberString is set to 1.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"In section 4 we evaluate these transformations with respect to projectivized dependency treebanks, and in section 5 they are applied to parser output.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"This feature has a linguistic justification.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"In total 13,976 phrases are assigned to sets of phrases, and the accuracy on our evaluation data ranges from 65 to 99%, depending on the domain and the size of the sets.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"We start with noun features since written Arabic contains a very high proportion of NPs.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"It is difficult to directly compare the Matsoukas et al results with ours, since our out-of-domain corpus is homogeneous; given heterogeneous training data, however, it would be trivial to include Matsoukas-style identity features in our instance-weighting model.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"7 Acknowledgements.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"The matching sentence pairs are then added to the IN corpus, and the system is re-trained.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"One may argue with these efforts on normalization, and ultimately their value should be assessed by assessing their impact on inter-annotator agreement.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"The domain is general politics, economics and science.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"The ATB segmentation scheme is one of many alternatives.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"However, to classify a token , while Borthwick uses tokens from to (from two tokens before to two tokens after ), we used only the tokens , , and . Even with local features alone, MENERGI outperforms MENE (Borthwick, 1999).","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"However, those methods need initial seeds, so the relation between entities has to be known in advance.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"Figure 1 reveals that an event that \u00e2\u0080\u009cdamaged\u00e2\u0080\u009d objects may also cause injuries; a disaster that \u00e2\u0080\u009coccurred\u00e2\u0080\u009d may be investigated to find its \u00e2\u0080\u009ccause\u00e2\u0080\u009d; a disaster may \u00e2\u0080\u009cwreak\u00e2\u0080\u009d havoc as it \u00e2\u0080\u009ccrosses\u00e2\u0080\u009d geographic regions; and vehicles that have a \u00e2\u0080\u009cdriver\u00e2\u0080\u009d may also \u00e2\u0080\u009ccarry\u00e2\u0080\u009d items.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"It can also be seen clearly in this plot that two of the Taiwan speakers cluster very closely together, and the third Tai\u00c2\u00ad wan speaker is also close in the most significant dimension (the x axis).","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"Covering the first uncovered position in the source sentence, we use the language model probability p(ej$; $).","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"The ATB annotation distinguishes between verbal and nominal readings of maSdar process nominals.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"(2009) also report results on English, but on the reduced 17 tag set, which is not comparable to ours).","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"It is the performance we could achieve if an omniscient observer told us which parser to pick for each of the sentences.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"In MSA, SVO usually appears in non-matrix clauses.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Out-of-domain test data is from the Project Syndicate web site, a compendium of political commentary.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"The resulting algorithm has a complexity of O(n!).","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"The way judgements are collected, human judges tend to use the scores to rank systems against each other.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"For statistics on this test set, refer to Figure 1.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"The system described in this paper is similar to the MENE system of (Borthwick, 1999).","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"Gabbard and Kulick (2008) show that there is significant attachment ambiguity associated with iDafa, which occurs in 84.3% of the trees in our development set.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Finally, quite a few hanzi are homographs, meaning that they may be pronounced in several different ways, and in extreme cases apparently represent different morphemes: The prenominal modifi\u00c2\u00ad cation marker eg deO is presumably a different morpheme from the second morpheme of \u00c2\u00a7eg mu4di4, even though they are written the same way.4 The second point, which will be relevant in the discussion of personal names in Section 4.4, relates to the internal structure of hanzi.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"(b) does the translation have the same meaning, including connotations?","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"These alignment models are similar to the concept of hidden Markov models (HMM) in speech recognition.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"2.1 Overview.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"When dropping the top and bottom 2.5% the remaining BLEU scores define the range of the confidence interval.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Again, we can compute average scores for all systems for the different language pairs (Figure 6).","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"State will ultimately be used as context in a subsequent query.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"Of these cases, 38 were temporal expressions (either a day of the week or month of the year).","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"Instead, we want to apply an inverse transformation to recover the underlying (nonprojective) dependency graph.","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"The last four columns in Table 3 show the distribution of nonprojective arcs with respect to the number of lifts required.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"information structure.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"It is not easy to make a clear definition of \u00e2\u0080\u009cparaphrase\u00e2\u0080\u009d.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"can expect famous names like Zhou Enlai's to be in many dictionaries, but names such as :fi lf;f; shi2jil-lin2, the name of the second author of this paper, will not be found in any dictionary.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Nonetheless there is no alternative to Reiche\u00e2\u0080\u0099s plan.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"4 Traditional Arabic linguistic theory treats both of these types as subcategories of noun \u00ef\u00bf\u00bd '.i . Figure 1: The Stanford parser (Klein and Manning, 2002) is unable to recover the verbal reading of the unvocalized surface form 0 an (Table 1).","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"(If the TF\/IDF score of that word is below a threshold, the phrase is discarded.)","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"We use two common techniques, hash tables and sorted arrays, describing each before the model that uses the technique.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Figure 2 shows a screenshot (which is of somewhat limited value, though, as color plays a major role in signalling the different statuses of the information).","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"At most one feature in this group will be set to 1.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"As a result, its POS tag needs to be induced in the \u201cNo LP\u201d case, while the 11A word level paired-t-test is significant at p < 0.01 for Danish, Greek, Italian, Portuguese, Spanish and Swedish, and p < 0.05 for Dutch. correct tag is available as a constraint feature in the \u201cWith LP\u201d case.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"(Again, the goal of also in structural features.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"Steedman (1986) considers Categorial Grammars in which both the operations of function application and composition may be used, and in which function can specify whether they take their arguments from their right or left.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"(Charniak et al., 1996).","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"Before describing the unsupervised case we first describe the supervised version of the algorithm: Input to the learning algorithm: n labeled examples of the form (xi, y\u201e). y, is the label of the ith example (given that there are k possible labels, y, is a member of y = {1 ... 0). xi is a set of mi features {x,1, Xi2 .","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"These 140 NE categories are designed by extending MUC\u00e2\u0080\u0099s 7 NE categories with finer sub-categories (such as Company, Institute, and Political Party for Organization; and Country, Province, and City for Location) and adding some new types of NE categories (Position Title, Product, Event, and Natural Object).","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"However, until such standards are universally adopted in evaluating Chinese segmenters, claims about performance in terms of simple measures like percent correct should be taken with a grain of salt; see, again, Wu and Fung (1994) for further arguments supporting this conclusion.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"Future extensions of the system might include: 1) An extended translation model, where we use more context to predict a source word.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"30 16.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"Having found (spelling, context) pairs in the parsed data, a number of features are extracted.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"However, those methods need initial seeds, so the relation between entities has to be known in advance.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"3.1 Lexicon Component.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"67 95.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"This is equivalent to the assumption used in probability estimation for na\u00efve Bayes classifiers, namely that the attribute values are conditionally independent when the target value is given.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"In this case, Maury Cooper is extracted.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"These knowledge sources determine whether the contexts surrounding an anaphor and antecedent are compatible.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"\"c' 0 + 0 \"0 ' \u00e2\u0080\u00a2 + a n t i g r e e d y x g r e e d y < > c u r r e n t m e t h o d o d i e t . o n l y \u00e2\u0080\u00a2 Taiwan 0 \u00c2\u00b7;; 0 c CD E i5 0\"' 9 9 \u00e2\u0080\u00a2 Mainland \u00e2\u0080\u00a2 \u00e2\u0080\u00a2 \u00e2\u0080\u00a2 \u00e2\u0080\u00a2 -0.30.20.1 0.0 0.1 0.2 Dimension 1 (62%) Figure 7 Classical metric multidimensional scaling of distance matrix, showing the two most significant dimensions.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"3.1 Word ReOrdering with Verbgroup.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"However, it can be noted that the results for the least informative encoding, Path, are almost comparable, while the third encoding, Head, gives substantially worse results for both data sets.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"However, this result is consistent with the results of ex\u00c2\u00ad periments discussed in Wu and Fung (1994).","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"There has been a lot of research on such lexical relations, along with the creation of resources such as WordNet.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"We have described grammar state splits that significantly improve parsing performance, catalogued parsing errors, and quantified the effect of segmentation errors.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"can expect famous names like Zhou Enlai's to be in many dictionaries, but names such as :fi lf;f; shi2jil-lin2, the name of the second author of this paper, will not be found in any dictionary.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Almost all annotators expressed their preference to move to a ranking-based evaluation in the future.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Thus, if one wants to segment words-for any purpose-from Chinese sentences, one faces a more difficult task than one does in English since one cannot use spacing as a guide.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Fortunately, there are only a few hundred hanzi that are particularly common in transliterations; indeed, the commonest ones, such as E. bal, m er3, and iij al are often clear indicators that a sequence of hanzi containing them is foreign: even a name like !:i*m xia4mi3-er3 'Shamir,' which is a legal Chi\u00c2\u00ad nese personal name, retains a foreign flavor because of liM.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"Using this heuristic, BABAR identifies existential definite NPs in the training corpus using our previous learning algorithm (Bean and Riloff, 1999) and resolves all occurrences of the same existential NP with each another.1 2.1.2 Syntactic Seeding BABAR also uses syntactic heuristics to identify anaphors and antecedents that can be easily resolved.","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"Memory-based classifiers for the experiments were created using TiMBL (Daelemans et al., 2003).","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"We briefly review it here for completeness.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"Ex: The government said it ...","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"Step 4.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"Following the setup of Johnson (2007), we use the whole of the Penn Treebank corpus for training and evaluation on English.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"However, when we pre- tag the input\u00e2\u0080\u0094as is recommended for English\u00e2\u0080\u0094 we notice a 0.57% F1 improvement.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"Our first model is GTplain, a PCFG learned from the treebank after removing all functional features from the syntactic categories.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Vocabulary lookup is a hash table mapping from word to vocabulary index.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"We also add an annotation for one-level iDafa (oneLevelIdafa) constructs since they make up more than 75% of the iDafa NPs in the ATB (Gabbard and Kulick, 2008).","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"As lower frequency examples include noise, we set a threshold that an NE category pair should appear at least 5 times to be considered and an NE instance pair should appear at least twice to be considered.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"This fact annoyed especially his dog...).","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"Thus the method makes the fairly strong assumption that the features can be partitioned into two types such that each type alone is sufficient for classification.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"The overall performance of the LTG system was outstanding, but the system consists of a sequence of many hand-coded rules and machine-learning modules.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"This is an iterative method that improves the estimation of the parameters at each iteration.","label":0},{"source_text":"Here we present two algorithms.","target_text":"(e.g., N.Y. would contribute this feature, IBM would not). nonalpha=x Appears if the spelling contains any characters other than upper or lower case letters.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"In (Reitter, Stede 2003) we went a different way and suggested URML5, an XML format for underspecifying rhetorical structure: a number of relations can be assigned instead of a single one, competing analyses can be represented with shared forests.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"The belief value that would have been assigned to the intersection of these sets is .60*.70=.42, but this belief has nowhere to go because the null set is not permissible in the model.7 So this probability mass (.42) has to be redistributed.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"We then use linguistic and annotation insights to develop a manually annotated grammar for Arabic (\u00c2\u00a74).","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"The NIST06 and NIST08 evaluation sets were used for testing.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"When extracting the vector t, used to compute the constraint feature from the graph, we tried three threshold values for r (see Eq.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Searching a probing hash table consists of hashing the key, indexing the corresponding bucket, and scanning buckets until a matching key is found or an empty bucket is encountered, in which case the key does not exist in the table.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"We carried out translation experiments in two different settings.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"While building a machine translation system is a serious undertaking, in future we hope to attract more newcomers to the field by keeping the barrier of entry as low as possible.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Systems that generally do worse than others will receive a negative one.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"(2010) and the posterior regular- ization HMM of Grac\u00c2\u00b8a et al.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"The availability of these resources guided our selection of foreign languages.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"The last affix in the list is the nominal plural f, men0.20 In the table are the (typical) classes of words to which the affix attaches, the number found in the test corpus by the method, the number correct (with a precision measure), and the number missed (with a recall measure).","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"During testing, it is possible that the classifier produces a sequence of inadmissible classes (e.g., person begin followed by location unique).","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"Thus, most broad-coverage parsers based on dependency grammar have been restricted to projective structures.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"IRSTLM\u2019s quantized variant is the inspiration for our quantized variant.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"Table 8a shows that the best model recovers SBAR at only 71.0% F1.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"The proof is given in (Tillmann, 2000).","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"For brevity we omit the segments from the analysis, and so analysis of the form \u201cfmnh\u201d as f\/REL mnh\/VB is represented simply as REL VB.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"The main disadvantage of manual evaluation is that it is time-consuming and thus too expensive to do frequently.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"In contrast to these approaches, our method directly incorporates these constraints into the structure of the model.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"Since the early days of statistical NLP, researchers have observed that a part-of-speech tag distribution exhibits \u00e2\u0080\u009cone tag per discourse\u00e2\u0080\u009d sparsity \u00e2\u0080\u0094 words are likely to select a single predominant tag in a corpus, even when several tags are possible.","label":0},{"source_text":"There is no global pruning.","target_text":"The proof is given in (Tillmann, 2000).","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"mark- ContainsVerb is especially effective for distinguishing root S nodes of equational sentences.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Let H be the set of hanzi, p be the set of pinyin syllables with tone marks, and P be the set of grammatical part-of-speech labels.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"5 70.1 58.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"While the linear precedence of segmental morphemes within a token is subject to constraints, the dominance relations among their mother and sister constituents is rather free.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"BABAR uses two methods to identify anaphors that can be easily and reliably resolved with their antecedent: lexical seeding and syntactic seeding.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Email: cls@bell-labs.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"(Blum and Mitchell 98) offer a promising formulation of redundancy, also prove some results about how the use of unlabeled examples can help classification, and suggest an objective function when training with unlabeled examples.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"52 15.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"In (b) is a plausible segmentation for this sentence; in (c) is an implausible segmentation.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"However, a recent study (Callison-Burch et al., 2006), pointed out that this correlation may not always be strong.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"The similarities become apparent when they are studied at the level of derivation structures: derivation nee sets of CFG's, HG's, TAG's, and MCTAG's are all local sets.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"We further thank Dr. J.-S.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"Preferences for constituent order (especially in languages with relatively free word order) often belong to this group.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"All annotations are done with specific tools and in XML; each layer has its own DTD.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"First we present the non-parametric version of parser switching, similarity switching: The intuition for this technique is that we can measure a similarity between parses by counting the constituents they have in common.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"the number of permutations carried out for the word reordering is given.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Other good classes include JADE and GOLD; other bad classes are DEATH and RAT.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"The supervised POS tagging accuracies (on this tagset) are shown in the last row of Table 2.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"Section 2 describes our baseline techniques for SMT adaptation, and section 3 describes the instance-weighting approach.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":", for A. T.&T. nonalpha.. .","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"5 \u00e2\u0080\u0098Underspecified Rhetorical Markup Language\u00e2\u0080\u0099 6 This confirms the figure given by (Schauer, Hahn.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"Consider for a set of constituents the isolated constituent precision parser metric, the portion of isolated constituents that are correctly hypothesized.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"In such cases we use the non-pruned lattice including all (possibly ungrammatical) segmentation, and let the statistics (including OOV) decide.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"3 Techniques for automatic vocalization have been studied (Zitouni et al., 2006; Habash and Rambow, 2007).","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"We can now compare this algorithm to that of (Yarowsky 95).","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"gaolgaolxing4xing4 'happily' In the particular form of A-not-A reduplication illustrated in (3a), the first syllable of the verb is copied, and the negative markerbu4 'not' is inserted between the copy and the full verb.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"Instead, we want to apply an inverse transformation to recover the underlying (nonprojective) dependency graph.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"A defining characteristic of MSA is the prevalence of discourse markers to connect and subordinate words and phrases (Ryding, 2005).","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"8 57.3 +F EA TS be st me dia n 50.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"Instead, we focused on phrases and set the frequency threshold to 2, and so were able to utilize a lot of phrases while minimizing noise.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"The weak hypothesis can abstain from predicting the label of an instance x by setting h(x) = 0.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"For each feature type f and tag t, a multinomial \u00cf\u0088tf is drawn from a symmetric Dirichlet distribution with concentration parameter \u00ce\u00b2.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"We asked participants to each judge 200\u2013300 sentences in terms of fluency and adequacy, the most commonly used manual evaluation metrics.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"First, from a large corpus, we extract all the NE instance pairs.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"For the remaining arcs, if the segment is in fact a known lexeme it is tagged as usual, but for the OOV arcs which are valid Hebrew entries lacking tags assignment, we assign all possible tags and then simulate a grammatical constraint.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"Systems that generally do better than others will receive a positive average normalizedjudgement per sentence.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"RandLM 0.2 (Talbot and Osborne, 2007) stores large-scale models in less memory using randomized data structures.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"Our oracles took advantage of the labeled treebanks: While we tried to minimize the number of free parameters in our model, there are a few hyperparameters that need to be set.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"Like CFG's, the choice is predetermined by a finite number of rules encapsulated in the grammar.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"In (b) is a plausible segmentation for this sentence; in (c) is an implausible segmentation.","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"The results are given in Table 4.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"The human judges were presented with the following definition of adequacy and fluency, but no additional instructions:","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Further, Maamouri and Bies (2004) argued that the English guidelines generalize well to other languages.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"of Arabic.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"Most of these groups follow a phrase-based statistical approach to machine translation.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"We first note that the accuracy results of our system are overall higher on their setup, on all measures, indicating that theirs may be an easier dataset.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"We first give a brief overview of boosting algorithms.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"Second, we treat the projected labels as features in an unsupervised model (\u00a75), rather than using them directly for supervised training.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"To explicitly handle the word reordering between words in source and target language, we use the concept of the so-called inverted alignments as given in (Ney et al., 2000).","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"In all figures, we present the per-sentence normalized judgements.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"For comparison to information-retrieval inspired baselines, eg (L\u00a8u et al., 2007), we select sentences from OUT using language model perplexities from IN.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"Trying to integrate constituent ordering and choice of referring expressions, (Chiarcos 2003) developed a numerical model of salience propagation that captures various factors of author\u00e2\u0080\u0099s intentions and of information structure for ordering sentences as well as smaller constituents, and picking appropriate referring expressions.10 Chiarcos used the PCC annotations of co-reference and information structure to compute his numerical models for salience projection across the generated texts.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"As discussed in Section 3, independent paths in tree sets, rather than the path complexity, may be crucial in characterizing semilinearity and polynomial time recognition.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"In this paper we have argued that Chinese word segmentation can be modeled ef\u00c2\u00ad fectively using weighted finite-state transducers.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"In this way we restrict the parameterization of a Language Original case English Danish Dutch German Spanish Swedish Portuguese 94.6 96.3 96.6 95.5 95.4 93.3 95.6 Table 1: Upper bound on tagging accuracy assuming each word type is assigned to majority POS tag.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"In our work, we demonstrate that using a simple na\u00c2\u00a8\u00c4\u00b1veBayes approach also yields substantial performance gains, without the associated training complexity.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"We include a constituent in our hypothesized parse if it appears in the output of a majority of the parsers.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"In our model there are no such hyper-parameters, and the performance is the result of truly joint disambiguation. sults.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"To optimize left-to-right queries, we extend state to store backoff information: where m is the minimal context from Section 4.1 and b is the backoff penalty.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"However, it is also mostly political content (even if not focused on the internal workings of the European Union) and opinion.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"Table 3 shows BABAR\u00e2\u0080\u0099s performance when the four contextual role knowledge sources are added.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"However, we note that the final conditional estimates p(s|t) from a given phrase table maximize the likelihood of joint empirical phrase pair counts over a word-aligned corpus.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Nonetheless, the results of the comparison with human judges demonstrates that there is mileage being gained by incorporating models of these types of words.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"Recently, combination techniques have been investigated for part of speech tagging with positive results (van Halteren et al., 1998; Brill and Wu, 1998).","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"One knowledge source, called WordSemCFSem, is analogous to CFLex: it checks whether the anaphor and candidate antecedent are substitutable for one another, but based on their semantic classes instead of the words themselves.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"The use of global features has improved the performance on MUC6 test data from 90.75% to 93.27% (27% reduction in errors), and the performance on MUC7 test data from 85.22% to 87.24% (14% reduction in errors).","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"For the English RST-annotated corpus that is made available via LDC, his corresponding result is 62%.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"The accuracy is calculated as the ratio of the number of paraphrases to the total number of phrases in the set.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"All three curves remain steep at the maximum training set size of 18818 trees.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"Because many foreign word types are not aligned to an English word (see Table 3), and we do not run label propagation on the foreign side, we expect the projected information to have less coverage.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Thus, we feel fairly confident that for the examples we have considered from Gan's study a solution can be incorporated, or at least approximated, within a finite-state framework.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"In (b) \u00e2\u0080\u009cthey\u00e2\u0080\u009d refers to the kidnapping victims, but in (c) \u00e2\u0080\u009cthey\u00e2\u0080\u009d refers to the armed men.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"In the natural disasters domain, agents are often forces of nature, such as hurricanes or wildfires.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":".., Wn ) (obs) P T : Tag assigns (T1 ,.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Table 9 shows that MADA produces a high quality segmentation, and that the effect of cascading segmentation errors on parsing is only 1.92% F1.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"Surprisingly, the non-parametric switching technique also exhibited robust behaviour in this situation.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"Here, an NE instance pair is any pair of NEs separated by at most 4 syntactic chunks; for example, \u00e2\u0080\u009cIBM plans to acquire Lotus\u00e2\u0080\u009d.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"We extend Subramanya et al.\u2019s intuitions to our bilingual setup.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"Our work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"The P (T |\u00cf\u0088) distribution, in English for instance, should have very low mass for the DT (determiner) tag, since determiners are a very small portion of the vocabulary.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"3.2 Results.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"2.4 Underspecified rhetorical structure.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"We collected around 300\u2013400 judgements per judgement type (adequacy or fluency), per system, per language pair.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Given the limited number of judgements we received, we did not try to evaluate this.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"In various dialects of Mandarin certain phonetic rules apply at the word.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"Each unlabeled pair (x1,i, x2,i) is represented as an edge between nodes corresponding to x1,i and X2,i in the graph.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"Although this property is not structural, it depends on the structural property that sentences can be built from a finite set of clauses of bounded structure as noted by Joshi (1983\/85).","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"Table 5 provides insight into the behavior of different models in terms of the tagging lexicon they generate.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"The treebank has two versions, v1.0 and v2.0, containing 5001 and 6501 sentences respectively.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"1,000 of these were picked at random, and labeled by hand to produce a test set.","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"We show how a datadriven deterministic dependency parser, in itself restricted to projective structures, can be combined with graph transformation techniques to produce non-projective structures.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Examples are given in Table 4.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"Across all languages, high performance can be attained by selecting a single tag per word type.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"In-domain Systran scores on this metric are lower than all statistical systems, even the ones that have much worse human scores.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"The second concerns the methods used (if any) to ex\u00c2\u00ad tend the lexicon beyond the static list of entries provided by the machine-readable dictionary upon which it is based.","label":0},{"source_text":"Here we present two algorithms.","target_text":"The label for a test example with features x is then defined as In this paper we define h(x, y) as the following function of counts seen in training data: Count(x,y) is the number of times feature x is seen with label y in training data, Count(x) = EyEy Count(x, y). a is a smoothing parameter, and k is the number of possible labels.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"As we have noted in Section 2, the general semantic class to which a hanzi belongs is often predictable from its semantic radical.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Feature selection is implemented using a feature cutoff: features seen less than a small count during training will not be used.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"This PCFG is incorporated into the Stanford Parser, a factored model that chooses a 1-best parse from the product of constituency and dependency parses.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"Automatic paraphrase discovery is an important but challenging task.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"Also, in Information Extraction (IE), in which the system tries to extract elements of some events (e.g. date and company names of a corporate merger event), several event instances from different news articles have to be aligned even if these are expressed differently.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"Fortunately, performance was stable across various values, and we were able to use the same hyperparameters for all languages.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"The resulting structural differences between tree- banks can account for relative differences in parsing performance.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"Month Names, Days of the Week, and Numbers: If is initCaps and is one of January, February, . . .","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"Ex: Mr. Cristiani, president of the country ...","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"Source sentence words are aligned with hypothesized target sentence words, where the choice of a new source word, which has not been aligned with a target word yet, is restricted1.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"The data structure was populated with 64-bit integers sampled uniformly without replacement.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"In MSA, SVO usually appears in non-matrix clauses.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"A crucial difference is that the number of parameters is greatly reduced as is the number of variables that are sampled during each iteration.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"For each terminal, the Leaf Ancestor metric extracts the shortest path to the root.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Vocabulary lookup is a hash table mapping from word to vocabulary index.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"9 50.2 +P RI OR be st me dia n 47.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"The family name set is restricted: there are a few hundred single-hanzi family names, and about ten double-hanzi ones.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Second, comparisons of different methods are not meaningful unless one can eval\u00c2\u00ad uate them on the same corpus.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"The 2(Yarowsky 95) describes the use of more sophisticated smoothing methods.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"For each pair of judges, consider one judge as the standard,.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"Step 3.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"We tabulate this increase in Table 3.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"If any of the tokens from to is in Person-Prefix- List, then another feature Person-Prefix is set to 1.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"When dropping the top and bottom 2.5% the remaining BLEU scores define the range of the confidence interval.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"The compact variant uses sorted arrays instead of hash tables within each node, saving some memory, but still stores full 64-bit pointers.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"In our third model GTppp we also add the distinction between general PPs and possessive PPs following Goldberg and Elhadad (2007).","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"The human judges were presented with the following definition of adequacy and fluency, but no additional instructions:","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"In many cases, inspection of either the spelling or context alone is sufficient to classify an example.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"We are very grateful to Tony Kroc.h, Michael Pails, Sunil Shende, and Mark Steedman for valuable discussions. formalisms.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"Even without features, but still using the tag prior, our median result is 52.0%, still significantly outperforming Grac\u00c2\u00b8a et al.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"This is motivated by taking \u03b2 po(s|t) to be the parameters of a Dirichlet prior on phrase probabilities, then maximizing posterior estimates p(s|t) given the IN corpus.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"If we were working with more than three parsers we could investigate minority constituents, those constituents that are suggested by at least one parser, but which the majority of the parsers do not suggest.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"8 Volume 1, English language, 19961997, Format version 1, correction level 0 An ap ho r T e r r o r i s m R e c Pr F D i s a s t e r s R e c Pr F De f. NP s Pro no uns .43 .79 .55 .50 .72 .59 .42 .91 .58 .42 .82 .56 Tot al .46 .76 .57 .42 .87 .57 Table 2: General Knowledge Sources Table 4: Individual Performance of KSs for Terrorism Table 3: General + Contextual Role Knowledge Sources larger MUC4 and Reuters corpora.9 4.2 Experiments.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"The similarities become apparent when they are studied at the level of derivation structures: derivation nee sets of CFG's, HG's, TAG's, and MCTAG's are all local sets.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"While many systems had similar performance, the results offer interesting insights, especially about the relative performance of statistical and rule-based systems.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"The manual evaluation of scoring translation on a graded scale from 1\u20135 seems to be very hard to perform.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"Clearly this poses a number of research challenges, though, such as the applicability of tag sets across different languages.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"Pairwise comparison is done using the sign test.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"We report results for the best and median hyperparameter settings obtained in this way.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"The normalized judgement per sentence is the raw judgement plus (0 minus average raw judgement for this judge on this sentence).","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"A typical situation is shown in Figure 1.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"Skipped (K): The translation of up to one word may be postponed . Verb (V): The translation of up to two words may be anticipated.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"In both cases the investigators were able to achieve significant improvements over the previous best tagging results.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"This differs from other implementations (Stolcke, 2002; Pauls and Klein, 2011) that use hash tables as nodes in a trie, as explained in the next section.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"However, dynamic programming can be used to find the shortest tour in exponential time, namely in O(n22n), using the algorithm by Held and Karp.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"We compare the manually annotated grammar, which we incorporate into the Stanford parser, to both the Berkeley (Petrov et al., 2006) and Bikel (Bikel, 2004) parsers.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"Consider the following sentences: (a) Jose Maria Martinez, Roberto Lisandy, and Dino Rossy, who were staying at a Tecun Uman hotel, were kidnapped by armed men who took them to an unknown place.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"All the sentences have been analyzed by our chunker and NE tag- ger.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"The distinctions in the ATB are linguistically justified, but complicate parsing.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"To this end, we construct a bilingual graph over word types to establish a connection between the two languages (\u00a73), and then use graph label propagation to project syntactic information from English to the foreign language (\u00a74).","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Different sentence structure and rich target language morphology are two reasons for this.","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"This may seem surprising, given the experiments reported in section 4, but the explanation is probably that the non-projective dependencies that can be recovered at all are of the simple kind that only requires a single lift, where the encoding of path information is often redundant.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"This work has been supported as part of the Verbmobil project (contract number 01 IV 601 A) by the German Federal Ministry of Education, Science, Research and Technology and as part of the Eutrans project (ESPRIT project number 30268) by the European Community.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Interestingly, Chang et al. report 80.67% recall and 91.87% precision on an 11,000 word corpus: seemingly, our system finds as many names as their system, but with four times as many false hits.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"na me =>1 ha nzi fa mi ly 1 ha nzi gi ve n 4.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"(2009), who also incorporate a sparsity constraint, but does via altering the model objective using posterior regularization.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"These results are achieved by training on the official MUC6 and MUC7 training data, which is much less training data than is used by other machine learning systems that worked on the MUC6 or MUC7 named entity task (Bikel et al., 1997; Bikel et al., 1999; Borth- wick, 1999).","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"tai2du2 'Taiwan Independence.'","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"The correct resolution in sentence (b) comes from knowledge that people who are kidnapped are often subsequently released.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"Therefore, we only score guess\/gold pairs with identical character yields, a condition that allows us to measure parsing, tagging, and segmentation accuracy by ignoring whitespace.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Trying to integrate constituent ordering and choice of referring expressions, (Chiarcos 2003) developed a numerical model of salience propagation that captures various factors of author\u00e2\u0080\u0099s intentions and of information structure for ordering sentences as well as smaller constituents, and picking appropriate referring expressions.10 Chiarcos used the PCC annotations of co-reference and information structure to compute his numerical models for salience projection across the generated texts.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"Although these existential NPs do not need a prior referent, they may occur multiple times in a document.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Second, comparisons of different methods are not meaningful unless one can eval\u00c2\u00ad uate them on the same corpus.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"In IE, creating the patterns which express the requested scenario, e.g. \u00e2\u0080\u009cmanagement succession\u00e2\u0080\u009d or \u00e2\u0080\u009ccorporate merger and acquisition\u00e2\u0080\u009d is regarded as the hardest task.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"The table shows that the lexicon tag frequency predicated by our full model are the closest to the gold standard.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"The choice of the particular newspaper was motivated by the fact that the language used in a regional daily is somewhat simpler than that of papers read nationwide.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"While the semantic aspect of radicals is by no means completely predictive, the semantic homogeneity of many classes is quite striking: for example 254 out of the 263 examples (97%) of the INSECT class listed by Wieger (1965, 77376) denote crawling or invertebrate animals; similarly 21 out of the 22 examples (95%) of the GHOST class (page 808) denote ghosts or spirits.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"Note that on some examples (around 2% of the test set) CoBoost abstained altogether; in these cases we labeled the test example with the baseline, organization, label.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"A search restriction especially useful for the translation direction from German to English is presented.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"Ex: Mr. Cristiani is the president ...","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"For seven out of eight languages a threshold of 0.2 gave the best results for our final model, which indicates that for languages without any validation set, r = 0.2 can be used.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"After the first step towards breadth had been taken with the PoS-tagging, RST annotation, and URML conversion of the entire corpus of 170 texts12 , emphasis shifted towards depth.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"To explore this tradeoff, we have performed experiments with three different encoding schemes (plus a baseline), which are described schematically in Table 1.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"After the first step towards breadth had been taken with the PoS-tagging, RST annotation, and URML conversion of the entire corpus of 170 texts12 , emphasis shifted towards depth.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Moreover, the Stanford parser achieves the most exact Leaf Ancestor matches and tagging accuracy that is only 0.1% below the Bikel model, which uses pre-tagged input.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"For the experiments in this paper we made a couple of additional modifications to the CoBoost algorithm.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"The last affix in the list is the nominal plural f, men0.20 In the table are the (typical) classes of words to which the affix attaches, the number found in the test corpus by the method, the number correct (with a precision measure), and the number missed (with a recall measure).","label":0},{"source_text":"Here we present two algorithms.","target_text":"We are currently exploring other methods that employ similar ideas and their formal properties.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"Again, we can compute average scores for all systems for the different language pairs (Figure 6).","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"The transitive closure of the dictionary in (a) is composed with Id(input) (b) to form the WFST (c).","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"Even when there is training data available in the domain of interest, there is often additional data from other domains that could in principle be used to improve performance.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"Additionally, it works about 3 times as fast as the IBM style search.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"Location list is processed into a list of unigrams and bigrams (e.g., New York).","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"However, while Dienes and Dubey recognize empty categories in a pre-processing step and only let the parser find their antecedents, we use the parser both to detect dislocated dependents and to predict either the type or the location of their syntactic head (or both) and use post-processing only to transform the graph in accordance with the parser\u2019s analysis.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"The rhetorical structure annotations of PCC have all been converted to URML.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Evaluation of Morphological Analysis.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Thus, if one wants to segment words-for any purpose-from Chinese sentences, one faces a more difficult task than one does in English since one cannot use spacing as a guide.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"0 57.3 51.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"Since each zi is a contiguous substring of the input (say ai,), and no two substrings overlap, we can represent zi by the pair of integers (i2, i2).","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"9 www.ling.unipotsdam.de\/sfb\/ Figure 2: Screenshot of Annis Linguistic Database 3.3 Symbolic and knowledge-based.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"As a (crude) approximation, we normalize the extraction patterns with respect to active and passive voice and label those extractions as agents or patients.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Annotation of syntactic structure for the core corpus has just begun.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"Judges where excluded from assessing the quality of MT systems that were submitted by their institution.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"As lower frequency examples include noise, we set a threshold that an NE category pair should appear at least 5 times to be considered and an NE instance pair should appear at least twice to be considered.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"Unfortunately, we have much less data to work with than with the automatic scores.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"This material is based upon work supported by the National Science Foundation Graduate Research Fellowship under Grant No.","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"However, the accuracy is considerably higher than previously reported results for robust non-projective parsing of Czech, with a best performance of 73% UAS (Holan, 2004).","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"If they knew that the first four words in a hypergraph node would never extend to the left and form a 5-gram, then three or even fewer words could be kept in the backward state.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"For fair comparison, we have tabulated all results with the size of training data used (Table 5 and Table 6).","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"We report the F1 value of both measures.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"The development of a na\u00efve Bayes classifier involves learning how much each parser should be trusted for the decisions it makes.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Mandarin exhibits several such processes, including A-not-A question formation, il\u00c2\u00ad lustrated in (3a), and adverbial reduplication, illustrated in (3b): 3.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"We show that noun-noun vs. discourse-level coordination ambiguity in Arabic is a significant source of parsing errors (Table 8c).","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Having explained the various layers of annotation in PCC, we now turn to the question what all this might be good for.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"10 Here we use the Good-Turing estimate (Baayen 1989; Church and Gale 1991), whereby the aggregate probability of previously unseen instances of a construction is estimated as ni\/N, where N is the total number of observed tokens and n1 is the number of types observed only once.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"7).","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Vocabulary lookup is a hash table mapping from word to vocabulary index.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"0 X u} \"' o; .2 X X><X X XX X X X X X X x X X X X X x X V X X X X .;t'*- XXX:OX X X X X X X 9 x X X XX XX X X X X X X X XXX:< X X>O<XX>!KXX XI<>< \u00c2\u00bbC X X XX :X: X X \"' X X XX >OO<X>D<XIK X X X X X X --XX\u00c2\u00bb: XXX X X\u00c2\u00bbC X X\u00c2\u00abX...C:XXX X Xll< X X ><XX>IIC:liiC:oiiiiCI--8!X:liiOC!I!S8K X X X 10 100 1000 10000 log(F)_base: R\"2=0.20 (p < 0.005) X 100000 Figure 6 Plot of log frequency of base noun, against log frequency of plural nouns.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"We present KenLM, a library that implements two data structures for efficient language model queries, reducing both time and costs.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"The latter arcs correspond to OOV words in English.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"To combat the proliferation of parsing edges, we prune the lattices according to a hand-constructed lexicon of 31 clitics listed in the ATB annotation guidelines (Maamouri et al., 2009a).","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"I = 1X21 N and N is a &quot;medium&quot; sized number so that it is feasible to collect 0(N) unlabeled examples.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"First, the model assumes independence between the first and second hanzi of a double given name.","label":0},{"source_text":"There is no global pruning.","target_text":"Using these states, we define partial hypothesis extensions, which are of the following type: (S0;C n fjg; j0) !","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"This approach leads to a search procedure with complexity O(E3 J4).","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"Adding the isolated constituents to our hypothesis parse could increase our expected recall, but in the cases we investigated it would invariably hurt our precision more than we would gain on recall.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"For instance, for out-ofdomain English-French, Systran has the best BLEU and manual scores.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"As we will see from Table 3, not much improvement is derived from this feature.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"We computed BLEU scores for each submission with a single reference translation.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"The algorithm can be viewed as heuristically optimizing an objective function suggested by (Blum and Mitchell 98); empirically it is shown to be quite successful in optimizing this criterion.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"In addition, as the CRF and PCFG look at similar sorts of information from within two inherently different models, they are far from independent and optimizing their product is meaningless.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"For parsing, this is a mistake, especially in the case of interrogatives.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Section 4.1 explained that state s is stored by applications with partial hypotheses to determine when they can be recombined.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Ltd., then organization will be more probable.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"The reason why we did not train with both MUC6 and MUC7 training data at the same time is because the task specifications for the two tasks are not identical.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"The hyperparameters \u00ce\u00b1 and \u00ce\u00b2 represent the concentration parameters of the token- and type-level components of the model respectively.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"The semantic agreement KS eliminates some candidates, but also provides positive evidence in one case: if the candidate and anaphor both have semantic tags human, company, date, or location that were assigned via NER or the manually labeled dictionary entries.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"This approach makes the training objective more complex by adding linear constraints proportional to the number of word types, which is rather prohibitive.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"The current work treats both segmental and super-segmental phenomena, yet we note that there may be more adequate ways to treat supersegmental phenomena assuming Word-Based morphology as we explore in (Tsarfaty and Goldberg, 2008).","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"We construct a mapping from all the space-delimited tokens seen in the training sentences to their corresponding analyses.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"of Tokens No.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"Our experiments consistently demonstrate that this model architecture yields substantial performance gains over more complex tagging counterparts.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"The sources of our dictionaries are listed in Table 2.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"However, it is also mostly political content (even if not focused on the internal workings of the European Union) and opinion.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"The treebank has two versions, v1.0 and v2.0, containing 5001 and 6501 sentences respectively.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"Here, \u00e2\u0080\u009cEG\u00e2\u0080\u009d represents \u00e2\u0080\u009cEastern Group Plc\u00e2\u0080\u009d.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"The maximum precision row is the upper bound on accuracy if we could pick exactly the correct constituents from among the constituents suggested by the three parsers.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"Recently, this topic has been getting more attention, as is evident from the Paraphrase Workshops in 2003 and 2004, driven by the needs of various NLP applications.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Figure 4 Input lattice (top) and two segmentations (bottom) of the sentence 'How do you say octopus in Japanese?'.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"One annotator suggested that this was the case for as much as 10% of our test sentences.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"However, there is a strong relationship between ni1s and the number of hanzi in the class.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Consequently, we implemented our own annotation tool ConAno in Java (Stede, Heintze 2004), which provides specifically the functionality needed for our purpose.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"We also collapse unary chains withidentical basic categories like NP \u00e2\u0086\u0092 NP.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"The type-level posterior term can be computed according to, P (Ti|W , T \u00e2\u0088\u0092i, \u00ce\u00b2) \u00e2\u0088\u009d Note that each round of sampling Ti variables takes time proportional to the size of the corpus, as with the standard token-level HMM.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"Out of those 15 links, 4 are errors, namely \u00e2\u0080\u009cbuy - pay\u00e2\u0080\u009d, \u00e2\u0080\u009cacquire - pay\u00e2\u0080\u009d, \u00e2\u0080\u009cpurchase - stake\u00e2\u0080\u009d \u00e2\u0080\u009cacquisition - stake\u00e2\u0080\u009d.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Table 1 shows four words \u00e2\u0080\u009c 0 Indeed NN Indeed Saddamwhose unvocalized surface forms 0 an are indistinguishable.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"(2009).","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"Dropping the conditioning on 0 for brevity, and letting \u00afc\u03bb(s, t) = c\u03bb(s, t) + yu(s|t), and \u00afc\u03bb(t) = 4Note that the probabilities in (7) need only be evaluated over the support of \u02dcp(s, t), which is quite small when this distribution is derived from a dev set.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Pairwise comparison: We can use the same method to assess the statistical significance of one system outperforming another.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"This procedure is repeated for T rounds while alternating between the two classifiers.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"We incur some additional memory cost due to storing state in each hypothesis, though this is minimal compared with the size of the model itself.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"Case and Zone of and : Similarly, if (or ) is initCaps, a feature (initCaps, zone) (or (initCaps, zone) ) is set to 1, etc. Token Information: This group consists of 10 features based on the string , as listed in Table 1.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Email: cls@bell-labs.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"(b) 89 :1 t& tal de cai2neng2 hen3 he DE talent very 'He has great talent' f.b ga ol hig h While the current algorithm correctly handles the (b) sentences, it fails to handle the (a) sentences, since it does not have enough information to know not to group the sequences.ma3lu4 and?]cai2neng2 respectively.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"The problem of &quot;noise&quot; items that do not fall into any of the three categories also needs to be addressed.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Finally, this effort is part of a much larger program that we are undertaking to develop stochastic finite-state methods for text analysis with applications to TIS and other areas; in the final section of this paper we will briefly discuss this larger program so as to situate the work discussed here in a broader context.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"In-domain Systran scores on this metric are lower than all statistical systems, even the ones that have much worse human scores.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"genitiveMark indicates recursive NPs with a indefinite nominal left daughter and an NP right daughter.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"Feature-based HMM Model (Berg- Kirkpatrick et al., 2010): The KM model uses a variety of orthographic features and employs the EM or LBFGS optimization algorithm; Posterior regulariation model (Grac\u00c2\u00b8a et al., 2009): The G10 model uses the posterior regular- ization approach to ensure tag sparsity constraint.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Chinese word segmentation can be viewed as a stochastic transduction problem.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"The English data comes from the WSJ portion of the Penn Treebank and the other languages from the training set of the CoNLL-X multilingual dependency parsing shared task.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"The authors acknowledge the support of the NSF (CAREER grant IIS0448168, and grant IIS 0904684).","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"The content does not necessarily reflect the views of the U.S. Government, and no official endorsement should be inferred.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"When a comparison against previous results requires additional pre-processing, we state it explicitly to allow for the reader to replicate the reported results.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"We evaluated the results based on two metrics.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Hash tables are a common sparse mapping technique used by SRILM\u2019s default and BerkeleyLM\u2019s hashed variant.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"For example, even if the contexts surrounding an anaphor and candidate match exactly, they are not coreferent if they have substantially different meanings 9 We would be happy to make our manually annotated test data available to others who also want to evaluate their coreference resolver on the MUC4 or Reuters collections.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"For each terminal, the Leaf Ancestor metric extracts the shortest path to the root.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"Lexical and Morphological Ambiguity The rich morphological processes for deriving Hebrew stems give rise to a high degree of ambiguity for Hebrew space-delimited tokens.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Annotators suggested that long sentences are almost impossible to judge.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"In the ATB, :: b asta\u00e2\u0080\u0099adah is tagged 48 times as a noun and 9 times as verbal noun.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"For brevity we omit the segments from the analysis, and so analysis of the form \u201cfmnh\u201d as f\/REL mnh\/VB is represented simply as REL VB.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Same for . In the case where the next token is a hyphen, then is also used as a feature: (init- Caps, ) is set to 1.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Therefore, we only score guess\/gold pairs with identical character yields, a condition that allows us to measure parsing, tagging, and segmentation accuracy by ignoring whitespace.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"They demonstrated this with the comparison of statistical systems against (a) manually post-edited MT output, and (b) a rule-based commercial system.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"The original tag set for the CoNLL-X Dutch data set consists of compounded tags that are used to tag multi-word units (MWUs) resulting in a tag set of over 300 tags.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"For example, syntactic decoders (Koehn et al., 2007; Dyer et al., 2010; Li et al., 2009) perform dynamic programming parametrized by both backward- and forward-looking state.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"3 68.4 49.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"The word joining is done on the basis of a likelihood criterion.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"Sometime, multiple words are needed, like \u00e2\u0080\u009cvice chairman\u00e2\u0080\u009d, \u00e2\u0080\u009cprime minister\u00e2\u0080\u009d or \u00e2\u0080\u009cpay for\u00e2\u0080\u009d (\u00e2\u0080\u009cpay\u00e2\u0080\u009d and \u00e2\u0080\u009cpay for\u00e2\u0080\u009d are different senses in the CC-domain).","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"Information from a sentence is sometimes insufficient to classify a name correctly.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"We showed in \u00c2\u00a72 that lexical ambiguity explains the underperformance of these categories.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Foreign names are usually transliterated using hanzi whose sequential pronunciation mimics the source language pronunciation of the name.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"This work was funded by NSF grant IRI-9502312.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"The set is then compared with the set generated from the Penn Treebank parse to determine the precision and recall.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"On one end of the spectrum are clustering approaches that assign a single POS tag to each word type (Schutze, 1995; Lamar et al., 2010).","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"We believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"Nonetheless there is no alternative to Reiche\u00e2\u0080\u0099s plan.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"we perform five runs with different random initialization of sampling state.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"Our final average POS tagging accuracy of 83.4% compares very favorably to the average accuracy of Berg-Kirkpatrick et al.\u2019s monolingual unsupervised state-of-the-art model (73.0%), and considerably bridges the gap to fully supervised POS tagging performance (96.6%).","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"The quasi-monotone search performs best in terms of both error rates mWER and SSER.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"In order to evaluate and advance this approach, it helps to feed into the knowledge base data that is already enriched with some of the desired information \u00e2\u0080\u0094 as in PCC.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"As we shall see, most of the linked sets are paraphrases.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Much confusion has been sown about Chinese writing by the use of the term ideograph, suggesting that hanzi somehow directly represent ideas.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"MADA uses an ensemble of SVMs to first re-rank the output of a deterministic morphological analyzer.","label":0},{"source_text":"Here we present two algorithms.","target_text":"Training under this model involves estimation of parameter values for P(y), P(m) and P(x I y).","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"Without HSPELL-pruning, our simpler grammars are somewhat lagging behind, but as the grammars improve the gap is bridged.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"The CoBoost algorithm just described is for the case where there are two labels: for the named entity task there are three labels, and in general it will be useful to generalize the CoBoost algorithm to the multiclass case.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"Due to many similarly performing systems, we are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"Matsoukas et al (2009) generalize it by learning weights on sentence pairs that are used when estimating relative-frequency phrase-pair probabilities.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"There is a (costless) transition between the NC node and f,.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"We measured recall (Rec), precision (Pr), and the F-measure (F) with recall and precision equally weighted.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"In our implementation, we make perhaps the simplest choice of weak hypothesis.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"splitPUNC restores the convention of the WSJ.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"4.1 Dictionary Representation.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"Although this feature helps, we encounter one consequence of variable word order.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"The development of automatic scoring methods is an open field of research.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"3 58.3 40.","label":0},{"source_text":"There is no global pruning.","target_text":"The German finite verbs 'bin' (second example) and 'konnten' (third example) are too far away from the personal pronouns 'ich' and 'Sie' (6 respectively 5 source sentence positions).","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"The compact variant uses sorted arrays instead of hash tables within each node, saving some memory, but still stores full 64-bit pointers.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"The best answer to this is: many research labs have very competitive systems whose performance is hard to tell apart.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"In IE, creating the patterns which express the requested scenario, e.g. \u00e2\u0080\u009cmanagement succession\u00e2\u0080\u009d or \u00e2\u0080\u009ccorporate merger and acquisition\u00e2\u0080\u009d is regarded as the hardest task.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"Out-of-domain test data is from the Project Syndicate web site, a compendium of political commentary.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Our evaluation includes both weighted and un- weighted lattices.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"One knowledge source, called WordSemCFSem, is analogous to CFLex: it checks whether the anaphor and candidate antecedent are substitutable for one another, but based on their semantic classes instead of the words themselves.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Note that Wang, Li, and Chang's.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Almost all annotators expressed their preference to move to a ranking-based evaluation in the future.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"constitute names, since we have only their segmentation, not the actual classification of the segmented words.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"For instance, in the recent IWSLT evaluation, first fluency annotations were solicited (while withholding the source sentence), and then adequacy annotations.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"We do not adapt the alignment procedure for generating the phrase table from which the TM distributions are derived.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"We train and test on the CoNLL-X training set.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"We present several variations for the lexical component P (T , W |\u00cf\u0088), each adding more complex pa rameterizations.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"In our second model GTvpi we also distinguished finite and non-finite verbs and VPs as 10Lattice parsing can be performed by special initialization of the chart in a CKY parser (Chappelier et al., 1999).","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"The rest of the paper is structured as follows.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"SRILM\u2019s compact variant has an incredibly expensive destructor, dwarfing the time it takes to perform translation, and so we also modified Moses to avoiding the destructor by calling exit instead of returning normally.","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"Pseudo-Projective Dependency Parsing","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"We received submissions from 14 groups from 11 institutions, as listed in Figure 2.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"This is especially bad with PROBING because it is based on hashing and performs random lookups, but it is not intended to be used in low-memory scenarios.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"For unaligned words, we set the tag to the most frequent tag in the corresponding treebank.","label":0},{"source_text":"There is no global pruning.","target_text":"Source sentence words are aligned with hypothesized target sentence words, where the choice of a new source word, which has not been aligned with a target word yet, is restricted1.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"For example, it is well-known that one can build a finite-state bigram (word) model by simply assigning a state Si to each word Wi in the vocabulary, and having (word) arcs leaving that state weighted such that for each Wj and corresponding arc aj leaving Si, the cost on aj is the bigram cost of WiWj- (Costs for unseen bigrams in such a scheme would typically be modeled with a special backoff state.)","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"shortest match at each point.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"Under certain conditions the constituent voting and na\u00efve Bayes constituent combination techniques are guaranteed to produce sets of constituents with no crossing brackets.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Virginia) and -sia are normally transliterated as fbSi!","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"Figure 2: Order in which source positions are visited for the example given in Fig.1.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"First, we evaluated BABAR using only the seven general knowledge sources.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"This is not completely surprising, since all systems use very similar technology.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"The function A : F \u2014* C maps from the language specific fine-grained tagset F to the coarser universal tagset C and is described in detail in \u00a76.2: Note that when tx(y) = 1 the feature value is 0 and has no effect on the model, while its value is \u2212oc when tx(y) = 0 and constrains the HMM\u2019s state space.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"To provide a thorough analysis, we evaluated three baselines and two oracles in addition to two variants of our graph-based approach.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"In this paper, we show that the maximum entropy framework is able to make use of global information directly, and achieves performance that is comparable to the best previous machine learning-based NERs on MUC6 and MUC7 test data.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"The use of global features has improved the performance on MUC6 test data from 90.75% to 93.27% (27% reduction in errors), and the performance on MUC7 test data from 85.22% to 87.24% (14% reduction in errors).","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"We have estimated the performance of IdentiFinder ' 99 at 200K words of training data from the graphs.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"na me =>1 ha nzi fa mi ly 1 ha nzi gi ve n 4.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"For example, in Information Retrieval (IR), we have to match a user\u00e2\u0080\u0099s query to the expressions in the desired documents, while in Question Answering (QA), we have to find the answer to the user\u00e2\u0080\u0099s question even if the formulation of the answer in the document is different from the question.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"We check, how likely only up to k = 20 better scores out of n = 100 would have been generated by two equal systems, using the binomial distribution: If p(0..k; n, p) < 0.05, or p(0..k; n, p) > 0.95 then we have a statistically significant difference between the systems.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"3 68.9 50.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"We evaluate our model on seven languages exhibiting substantial syntactic variation.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Since Dane\u00cb\u0087s\u00e2\u0080\u0099 proposals of \u00e2\u0080\u0098thematic development patterns\u00e2\u0080\u0099, a few suggestions have been made as to the existence of a level of discourse structure that would predict the information structure of sentences within texts.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"The most accurate characterization of Chinese writing is that it is morphosyllabic (DeFrancis 1984): each hanzi represents one morpheme lexically and semantically, and one syllable phonologi\u00c2\u00ad cally.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"The work reported here is closely related to [Ha- segawa et al. 04].","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"The second stage links sets which involve the same pairs of individual NEs.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"This material is based upon work supported by the National Science Foundation Graduate Research Fellowship under Grant No.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"Secondly, for all our models we provide better fine- and coarse-grained POS-tagging accuracy, and all pruned models outperform the Oracle results reported by them.12 In terms of syntactic disambiguation, even the simplest grammar pruned with HSPELL outperforms their non-Oracle results.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"This design leads to a significant reduction in the computational complexity of training and inference.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"each word in the lexicon whether or not each string is actually an instance of the word in question.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"The manual scores are averages over the raw unnormalized scores.","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"Although the best published results for the Collins parser is 80% UAS (Collins, 1999), this parser reaches 82% when trained on the entire training data set, and an adapted version of Charniak\u2019s parser (Charniak, 2000) performs at 84% (Jan Haji\u02c7c, pers. comm.).","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"To show that the derivation trees of any grammar in LCFRS is a local set, we can rewrite the annotated derivation trees such that every node is labelled by a pair to include the composition operations.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"One key issue here is to seek a discourse-based model of information structure.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"It is closer to the smaller value of precision and recall when there is a large skew in their values.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"TIS systems in general need to do more than simply compute the.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"For example, the company may refer to Company X in one paragraph and Company Y in another.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"While Berg-Kirkpatrick et al.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"However, lazy mapping is generally slow because queries against uncached pages must wait for the disk.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"This has solutions: where pI(s|t) is derived from the IN corpus using relative-frequency estimates, and po(s|t) is an instance-weighted model derived from the OUT corpus.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"For the LM, adaptive weights are set as follows: where \u03b1 is a weight vector containing an element \u03b1i for each domain (just IN and OUT in our case), pi are the corresponding domain-specific models, and \u02dcp(w, h) is an empirical distribution from a targetlanguage training corpus\u2014we used the IN dev set for this.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"0 68.4 48.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"In Chinese text, individual characters of the script, to which we shall refer by their traditional name of hanzi,Z are written one after another with no intervening spaces; a Chinese sentence is shown in Figure 1.3 Partly as a result of this, the notion \"word\" has never played a role in Chinese philological tradition, and the idea that Chinese lacks any\u00c2\u00ad thing analogous to words in European languages has been prevalent among Western sinologists; see DeFrancis (1984).","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"This step ensures that the most frequent terms for each domain are labeled (in case some of them are not in WordNet) and labeled with the sense most appropriate for the domain.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"Here, the pruning threshold t0 = 10:0 is used.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"Also there are cases where one of the two NEs belong to a phrase outside of the relation.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"For each label (Per s on, organization and Location), take the n contextual rules with the highest value of Count' (x) whose unsmoothed3 strength is above some threshold pmin.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"(Blum and Mitchell 98) describe learning in the following situation: X = X1 X X2 where X1 and X2 correspond to two different &quot;views&quot; of an example.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"Each xt E 2x is the set of features constituting the ith example.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"Feature selection is implemented using a feature cutoff: features seen less than a small count during training will not be used.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"The problem can be represented as a graph with 2N vertices corresponding to the members of X1 and X2.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"In this section, we describe the seven general knowledge sources and explain how the DempsterShafer model makes resolutions.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"Two common cases are the attribu tive adjective and the process nominal _; maSdar, which can have a verbal reading.4 At tributive adjectives are hard because they are or- thographically identical to nominals; they are inflected for gender, number, case, and definiteness.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"In the named entity task, X1 might be the instance space for the spelling features, X2 might be the instance space for the contextual features.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"Supervised part-of-speech (POS) taggers, for example, approach the level of inter-annotator agreement (Shen et al., 2007, 97.3% accuracy for English).","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"They considered a semi-supervised POS tagging scenario and showed that one can use a graph over trigram types, and edge weights based on distributional similarity, to improve a supervised conditional random field tagger.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"As a partial solution, for pairs of hanzi that co-occur sufficiently often in our namelists, we use the estimated bigram cost, rather than the independence-based cost.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"There are simply not enough votes remaining to allow any of the crossing structures to enter the hypothesized constituent set.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Purely statistical approaches have not been very popular, and so far as we are aware earlier work by Sproat and Shih (1990) is the only published instance of such an approach.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"It uses a maximum entropy framework and classifies each word given its features.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"(Carlson, Marcu 2001) responded to this situation with relatively precise (and therefore long!)","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"To eliminate such sequences, we define a transition probability between word classes to be equal to 1 if the sequence is admissible, and 0 otherwise.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"0.2 0.1 0.0 -0.1 25 26 27 28 29 30 31 32 -0.2 -0.3 \u2022systran \u2022 ntt 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 20 21 22 23 24 25 26 Fluency Fluency \u2022systran \u2022nrc rali 25 26 27 28 29 30 31 32 0.2 0.1 0.0 -0.1 -0.2 -0.3 cme p \ufffd 20 21 22 23 24 25 26 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 Figure 14: Correlation between manual and automatic scores for English-French 119 In Domain Out of Domain \u2022upv Adequacy -0.9 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 \u2022upv 23 24 25 26 27 28 29 30 31 32 \u2022upc-mr \u2022utd \u2022upc-jmc \u2022uedin-birch \u2022ntt \u2022rali \u2022uedin-birch 16 17 18 19 20 21 22 23 24 25 26 27 Adequacy \u2022upc-mr 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 -0.9 -1.0 -1.1 English-Spanish Fluency \u2022ntt \u2022nrc \u2022rali \u2022uedin-birch -0.2 -0.3 -0.5 \u2022upv 16 17 18 19 20 21 22 23 24 25 26 27 -0.4 nr \u2022 rali Fluency -0.4 \u2022upc-mr utd \u2022upc-jmc -0.5 -0.6 \u2022upv 23 24 25 26 27 28 29 30 31 32 0.2 0.1 -0.0 -0.1 -0.2 -0.3 0.3 0.2 0.1 -0.0 -0.1 -0.6 -0.7 Figure 15: Correlation between manual and automatic scores for English-Spanish 120 English-German In Domain Out of Domain Adequacy Adequacy 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 \u2022upv -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 -0.9 \u2022upv 0.5 0.4 \u2022systran \u2022upc-mr \u2022 \u2022rali 0.3 \u2022ntt 0.2 0.1 -0.0 -0.1 \u2022systran \u2022upc-mr -0.9 9 10 11 12 13 14 15 16 17 18 19 6 7 8 9 10 11 Fluency 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 \u2022upv -0.5 \u2022upv \u2022systran \u2022upc-mr \u2022 Fluency 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 \u2022systran \u2022ntt","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"However, it is desirable if we can separate them.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"123 examples fell into the noise category.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"2 62.6 45.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"The difference between the featureless model (+PRIOR) and our full model (+FEATS) is 13.6% and 7.7% average error reduction on best and median settings respectively.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"The use of global features has improved the performance on MUC6 test data from 90.75% to 93.27% (27% reduction in errors), and the performance on MUC7 test data from 85.22% to 87.24% (14% reduction in errors).","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"The","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Indeed, as we shall show in Section 5, even human judges differ when presented with the task of segmenting a text into words, so a definition of the criteria used to determine that a given segmentation is correct is crucial before one can interpret such measures.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"We dropped, however, one of the languages, Finnish, partly to keep the number of tracks manageable, partly because we assumed that it would be hard to find enough Finnish speakers for the manual evaluation.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"The web-based Annis imports data in a variety of XML formats and tagsets and displays it in a tier-orientedway (optionally, trees can be drawn more ele gantly in a separate window).","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"We also mark all nodes that dominate an SVO configuration (containsSVO).","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Nonetheless, the prospect of a network of annotated discourse resources seems particularly promising if not only a single annotation layer is used but a whole variety of them, so that a systematic search for correlations between them becomes possible, which in turn can lead to more explanatory models of discourse structure.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"Another way to view the judgements is that they are less quality judgements of machine translation systems per se, but rankings of machine translation systems.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Given a sorted array A, these other packages use binary search to find keys in O(log |A|) time.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"Consider for a set of constituents the isolated constituent precision parser metric, the portion of isolated constituents that are correctly hypothesized.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"5 64.7 42.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"Pairwise comparison is done using the sign test.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"The Bayes models were able to achieve significantly higher precision than their non-parametric counterparts.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"Step 2.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"Fort= 1,...,T:","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"Automatic paraphrase discovery is an important but challenging task.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"Table 4 shows how much the Bayes switching technique uses each of the parsers on the test set.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"\u2022 Similarly, when the na\u00efve Bayes classifier is configured such that the constituents require estimated probabilities strictly larger than 0.5 to be accepted, there is not enough probability mass remaining on crossing brackets for them to be included in the hypothesis.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"It was also proposed to allow annotators to skip sentences that they are unable to judge.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"The substructures that are unanimously hypothesized by the parsers should be preserved after combination, and the combination technique should not foolishly create substructures for which there is no supporting evidence.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"The code is opensource, has minimal dependencies, and offers both C++ and Java interfaces for integration.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"5 Related Work.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"We used this data to build an unpruned ARPA file with IRSTLM\u2019s improved-kneser-ney option and the default three pieces.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"We now turn our attention to the recognition of string languages generated by these formalisms (LCFRL's).","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"vierten 12.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"In this way, the method reported on here will necessarily be similar to a greedy method, though of course not identical.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"The role that each noun phrase plays in the kidnapping event is key to distinguishing these cases.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Vocabulary lookup is a sorted array of 64-bit word hashes.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Nonstochastic lexical-knowledge-based approaches have been much more numer\u00c2\u00ad ous.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"In the named entity task, X1 might be the instance space for the spelling features, X2 might be the instance space for the contextual features.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"This locally normalized log-linear model can look at various aspects of the observation x, incorporating overlapping features of the observation.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"For example, management succession systems must distinguish between a person who is fired and a person who is hired.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Along with locking and background kernel operations such as prefaulting, this explains why wall time is not one-eighth that of the single-threaded case. aLossy compression with the same weights. bLossy compression with retuned weights. the non-lossy options.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"For each pair of judges, consider one judge as the standard,.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"We make the assumption that for each example, both xi,. and x2,2 alone are sufficient to determine the label yi.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"To lower the barrier of entrance to the competition, we provided a complete baseline MT system, along with data resources.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"While Berg-Kirkpatrick et al.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"We have not explored this strategy.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"By taking the ratio of matching n-grams to the total number of n-grams in the system output, we obtain the precision pn for each n-gram order n. These values for n-gram precision are combined into a BLEU score: The formula for the BLEU metric also includes a brevity penalty for too short output, which is based on the total number of words in the system output c and in the reference r. BLEU is sensitive to tokenization.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Conversely, the lattice parser requires no linguistic resources and produces segmentations of comparable quality.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"The manual scores are averages over the raw unnormalized scores.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"BABAR then computes statistics over the training examples measuring the frequency with which extraction patterns and noun phrases co-occur in coreference resolutions.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"Several coreference resolvers have used supervised learning techniques, such as decision trees and rule learners (Aone and Bennett, 1995; McCarthy and Lehnert, 1995; Ng and Cardie, 2002; Soon et al., 2001).","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"The segmenter handles the grouping of hanzi into words and outputs word pronunciations, with default pronunciations for hanzi it cannot group; we focus here primarily on the system's ability to segment text appropriately (rather than on its pronunciation abilities).","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Much confusion has been sown about Chinese writing by the use of the term ideograph, suggesting that hanzi somehow directly represent ideas.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"If the expression is longer or complicated (like \u00e2\u0080\u009cA buys B\u00e2\u0080\u009d and \u00e2\u0080\u009cA\u00e2\u0080\u0099s purchase of B\u00e2\u0080\u009d), it is called \u00e2\u0080\u009cparaphrase\u00e2\u0080\u009d, i.e. a set of phrases which express the same thing or event.","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"With respect to exact match, the improvement is even more noticeable, which shows quite clearly that even if non-projective dependencies are rare on the token level, they are nevertheless important for getting the global syntactic structure correct.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"This measure has the advantage of being completely automatic.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"The rewrite rules and the definition of the composition operations may be stored in the finite state control since G uses a finite number of them.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"2.2 Syntactic structure.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Therefore in cases where the segmentation is identical between the two systems we assume that tagging is also identical.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Terrorism systems must distinguish between people who perpetrate a crime and people who are victims of a crime.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"42 nator, the N31s can be measured well by counting, and we replace the expectation by the observation.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"This distributional sparsity of syntactic tags is not unique to English 1 The source code for the work presented in this paper is available at http:\/\/groups.csail.mit.edu\/rbg\/code\/typetagging\/.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"The highestorder N-gram array omits backoff and the index, since these are not applicable.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"11 www.ling.unipotsdam.de\/sfb\/projekt a3.php 12 This step was carried out in the course of the diploma thesis work of David Reitter (2003), which de serves special mention here.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"In Section 6 we dis\u00c2\u00ad cuss other issues relating to how higher-order language models could be incorporated into the model.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"Because all English vertices are going to be labeled, we do not need to disambiguate them by embedding them in trigrams.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"To initialize the graph we tag the English side of the parallel text using a supervised model.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"In this paper, we make a simplifying assumption of one-tag-per-word.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Instead, the designs of the various annotation layers and the actual annotation work are results of a series of diploma theses, of students\u00e2\u0080\u0099 work in course projects, and to some extent of paid assistentships.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"This annotation choice weakens splitIN.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"It can be shown that the path set of the tree set generated by a TAG G is a context-free language.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"(2010) and the posterior regular- ization HMM of Grac\u00c2\u00b8a et al.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"(5) and ht into Equ.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"In application settings, this may be a profitable strategy.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Evalb, the standard parsing metric, is biased toward such corpora (Sampson and Babarczy, 2003).","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"The matching sentence pairs are then added to the IN corpus, and the system is re-trained.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"2 62.6 45.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"These systems rely on a training corpus that has been manually annotated with coreference links.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"The sequence of states needed to carry out the word reordering example in Fig.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"For a corpus C, let M be the set of tuples \u00e2\u0088\u0097n, l), where n is an n-gram with bracketing label l. If any n appears 6 Generative parsing performance is known to deteriorate with sentence length.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"However, while Dienes and Dubey recognize empty categories in a pre-processing step and only let the parser find their antecedents, we use the parser both to detect dislocated dependents and to predict either the type or the location of their syntactic head (or both) and use post-processing only to transform the graph in accordance with the parser\u2019s analysis.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"In the third and final scheme, denoted Path, we keep the extra infor2Note that this is a baseline for the parsing experiment only (Experiment 2).","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"The Verbmobil task is an appointment scheduling task.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"Suffixes and Prefixes: This group contains only two features: Corporate-Suffix and Person-Prefix.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"1 55.8 38.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Entries landing in the same bucket are said to collide.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"The easiest language pair according to BLEU (English-French: 28.33) received worse manual scores than the hardest (English-German: 14.01).","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"The joint morphological and syntactic hypothesis was first discussed in (Tsarfaty, 2006; Tsarfaty and Sima\u2019an, 2004) and empirically explored in (Tsarfaty, 2006).","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Each model was able to produce hypotheses for all input sentences.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"It was also proposed to allow annotators to skip sentences that they are unable to judge.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"We describe a novel approach for inducing unsupervised part-of-speech taggers for languages that have no labeled training data, but have translated text in a resource-rich language.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Several methods exist to handle collisions; we use linear probing because it has less memory overhead when entries are small.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"In Table 1 we see with very few exceptions that the isolated constituent precision is less than 0.5 when we use the constituent label as a feature.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"Hence, s < k. But by addition of the votes on the two parses, s > 2N-11> k, a contradiction.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"The scores and confidence intervals are detailed first in the Figures 7\u201310 in table form (including ranks), and then in graphical form in Figures 11\u201316.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Since different judges judged different systems (recall that judges were excluded to judge system output from their own institution), we normalized the scores.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"For example, suppose the current model assigns a belief value of .60 to {A, B}, meaning that it is 60% sure that the correct hypothesis is either A or B. Then new evidence arrives with a belief value of .70 assigned 5 Initially there are no competing hypotheses because all hypotheses are included in \u00ce\u00b8 by definition.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Clearly the percentage of productively formed words is quite small (for this particular corpus), meaning that dictionary entries are covering most of the 15 GR is .73 or 96%..","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"Sentences and systems were randomly selected and randomly shuffled for presentation.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"We can only compare with Grac\u00c2\u00b8a et al.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"Two strings are letter equivalent if they contain equal number of occurrences of each terminal symbol, and two languages are letter equivalent if every string in one language is letter equivalent to a string in the other language and vice-versa.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"Across all languages, +PRIOR consistently outperforms 1TW, reducing error on average by 9.1% and 5.9% on best and median settings respectively.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"More recently, (Riloff and Jones 99) describe a method they term &quot;mutual bootstrapping&quot; for simultaneously constructing a lexicon and contextual extraction patterns.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"5.1 Data Sets.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"Thus, the ATM has no more than 6km&quot; + 1 work tapes, where km&quot; is the maximum number of substrings spanned by a derived structure.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Table 8b shows that verbal nouns are the hardest pre-terminal categories to identify.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Up to now, most IE researchers have been creating paraphrase knowledge (or IE patterns) by hand and for specific tasks.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"Hence, the different averages of manual scores for the different language pairs reflect the behaviour of the judges, not the quality of the systems on different language pairs.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"This WFST is then summed with the WFST implementing the dictionary and morphological rules, and the transitive closure of the resulting transducer is computed; see Pereira, Riley, and Sproat (1994) for an explanation of the notion of summing WFSTs.12 Conceptual Improvements over Chang et al.'s Model.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"We start with noun features since written Arabic contains a very high proportion of NPs.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"La ng ua ge 1T W + P RI O R + F E A T S E ng lis h D a ni s h D u tc h G e r m a n P or tu g u e s e S p a ni s h S w e di s h 2 1.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"In the denomi 11 We have two such lists, one containing about 17,000 full names, and another containing frequencies of.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"We carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"The Verbmobil task is an appointment scheduling task.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"Search CPU time mWER SSER Method [sec] [%] [%] MonS 0:9 42:0 30:5 QmS 10:6 34:4 23:8 IbmS 28:6 38:2 26:2 4.2 Performance Measures.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"For other languages, we use the CoNLL-X multilingual dependency parsing shared task corpora (Buchholz and Marsi, 2006) which include gold POS tags (used for evaluation).","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"The input for the segmentation task is however highly ambiguous for Semitic languages, and surface forms (tokens) may admit multiple possible analyses as in (BarHaim et al., 2007; Adler and Elhadad, 2006).","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"The interdependence between fb or 1\/!i, and 5:2 is not captured by our model, but this could easily be remedied.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Less frequently studied is the interplay among language, annotation choices, and parsing model design (Levy and Manning, 2003; Ku\u00c2\u00a8 bler, 2005).","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"Tsarfaty and Sima\u2019an (2007) have reported state-of-the-art results on Hebrew unlexicalized parsing (74.41%) albeit assuming oracle morphological segmentation.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"This suggests that different types of anaphora may warrant different treatment: definite NP resolution may depend more on lexical semantics, while pronoun resolution may depend more on contextual semantics.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"Hence, the different averages of manual scores for the different language pairs reflect the behaviour of the judges, not the quality of the systems on different language pairs.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"This limitation is the obstacle to making the technology \u00e2\u0080\u009copen domain\u00e2\u0080\u009d.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"This may be the sign of a maturing research environment.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"Reference resolution involves finding words that co-refer to the same entity.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"We present two algorithms.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"If any of the tokens from to is in Person-Prefix- List, then another feature Person-Prefix is set to 1.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"We propose a limit of 70 words for Arabic parsing evaluations.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"BABAR then computes statistics over the training examples measuring the frequency with which extraction patterns and noun phrases co-occur in coreference resolutions.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"21 In Chinese, numerals and demonstratives cannot modify nouns directly, and must be accompanied by.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"The first four affixes are so-called resultative affixes: they denote some prop\u00c2\u00ad erty of the resultant state of a verb, as in E7 wang4bu4-liao3 (forget-not-attain) 'cannot forget.'","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"Across all languages, +PRIOR consistently outperforms 1TW, reducing error on average by 9.1% and 5.9% on best and median settings respectively.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"In Figure 4 we show an example of variation between the parsing models.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"Again, the idea is that having a picture of syntax, co-reference, and sentence-internal information structure at one\u00e2\u0080\u0099s disposal should aid in finding models of discourse structure that are more explanatory and can be empirically supported.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Evaluation of the Segmentation as a Whole.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"There are two differences between this method and the DL-CoTrain algorithm: spelling and contextual features, alternating between labeling and learning with the two types of features.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"Also, the argument has been made that machine translation performance should be evaluated via task-based evaluation metrics, i.e. how much it assists performing a useful task, such as supporting human translators or aiding the analysis of texts.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"Compared to last year\u2019s shared task, the participants represent more long-term research efforts.","label":0},{"source_text":"Here we present two algorithms.","target_text":"The key point is that the second constraint can be remarkably powerful in reducing the complexity of the learning problem.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Compared with the widely- SRILM, our is 2.4 times as fast while using 57% of the mem- The structure is a trie with bit-level packing, sorted records, interpolation search, and optional quantization aimed lower memory consumption. simultaneously uses less memory than the smallest lossless baseline and less CPU than the baseline.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"In previous work (Bean and Riloff, 1999), we developed an unsupervised learning algorithm that automatically recognizes definite NPs that are existential without syntactic modification because their meaning is universally understood.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"The perplexity for the trigram language model used is 26:5.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Interpolation search has a more expensive pivot but performs less pivoting and reads, so it is slow on small data and faster on large data.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"Co-occurrences among the particles themselves are subject to further syntactic and lexical constraints relative to the stem.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"We confirm the finding by Callison-Burch et al. (2006) that the rule-based system of Systran is not adequately appreciated by BLEU.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Recently, lattices have been used successfully in the parsing of Hebrew (Tsarfaty, 2006; Cohen and Smith, 2007), a Semitic language with similar properties to Arabic.","label":0},{"source_text":"Here we present two algorithms.","target_text":"In the cotraining case, (Blum and Mitchell 98) argue that the task should be to induce functions Ii and f2 such that So Ii and 12 must (1) correctly classify the labeled examples, and (2) must agree with each other on the unlabeled examples.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"Before presenting our results, we describe the datasets that we used, as well as two baselines.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"We further thank Dr. J.-S.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"The final model tions.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"However, dynamic programming can be used to find the shortest tour in exponential time, namely in O(n22n), using the algorithm by Held and Karp.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"The table shows that the lexicon tag frequency predicated by our full model are the closest to the gold standard.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"If the semantic expectations do not intersect, then we know that the case- frames extract mutually exclusive types of noun phrases.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"Precision is the portion of hypothesized constituents that are correct and recall is the portion of the Treebank constituents that are hypothesized.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"Sc opi ng filters candidate if outside the anaphor\u00e2\u0080\u0099s scope.","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"The first setting uses the European Medicines Agency (EMEA) corpus (Tiedemann, 2009) as IN, and the Europarl (EP) corpus (www.statmt.org\/europarl) as OUT, for English\/French translation in both directions.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"An anti-greedy algorithm, AG: instead of the longest match, take the.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Figure 1 provides some statistics about this corpus.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"A consequence of the ability to generate tree sets with this property is that CC's under this definition can generate the following language which can not be generated by either TAG's or HG's.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"This approach makes the training objective more complex by adding linear constraints proportional to the number of word types, which is rather prohibitive.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Hence, the different averages of manual scores for the different language pairs reflect the behaviour of the judges, not the quality of the systems on different language pairs.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"gaolxing4 'happy'=> F.i'JF.i'JJI!JI!","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"Based on these high-confidence alignments we can extract tuples of the form [u H v], where u is a foreign trigram type, whose middle word aligns to an English word type v. Our bilingual similarity function then sets the edge weights in proportion to these tuple counts.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"With the exception of the Dutch data set, no other processing is performed on the annotated tags.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"The number of NE instance pairs used in their experiment is less than half of our method.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"For simplicity, we assume that OUT is homogeneous.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"We can define derivation trees inductively on the length of the derivation of a tree 1.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"Step 1.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"A defining characteristic of MSA is the prevalence of discourse markers to connect and subordinate words and phrases (Ryding, 2005).","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"Finally, Section 5 explains how BABAR relates to previous work, and Section 6 summarizes our conclusions.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"The second row represents the performance of the median hyperparameter setting.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"One is that smaller sets sometime have meaningless keywords, like \u00e2\u0080\u009cstrength\u00e2\u0080\u009d or \u00e2\u0080\u009cadd\u00e2\u0080\u009d in the CC-domain, or \u00e2\u0080\u009ccompare\u00e2\u0080\u009d in the PC-domain.","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"We will also directly compare with a baseline similar to the Matsoukas et al approach in order to measure the benefit from weighting phrase pairs (or ngrams) rather than full sentences.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"Training and testing is based on the Europarl corpus.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"Hence, there is a total of 29 classes (7 name classes 4 sub-classes 1 not-a-name class).","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"Compared to the EMEA\/EP setting, the two domains in the NIST setting are less homogeneous and more similar to each other; there is also considerably more IN text available.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"In the small window on the left, search queries can be entered, here one for an NP that has been annotated on the co-reference layer as bridging.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"With a good hash function, collisions of the full 64bit hash are exceedingly rare: one in 266 billion queries for our baseline model will falsely find a key not present.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"We have described two data structures for language modeling that achieve substantial reductions in time and memory cost.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"If we were working with more than three parsers we could investigate minority constituents, those constituents that are suggested by at least one parser, but which the majority of the parsers do not suggest.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"16 As one reviewer points out, one problem with the unigram model chosen here is that there is still a. tendency to pick a segmentation containing fewer words.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"In this paper we present a stochastic finite-state model for segmenting Chinese text into words, both words found in a (static) lexicon as well as words derived via the above-mentioned productive processes.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"29 \u00e2\u0080\u0094 95.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"Consider the case where IX].","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"We do not show the numbers for the Bayes models in Table 2 because the parameters involved were established using this set.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Saving state allows our code to walk the data structure exactly once per query.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"We quantify error categories in both evaluation settings.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"One may argue with these efforts on normalization, and ultimately their value should be assessed by assessing their impact on inter-annotator agreement.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"We observe similar trends when using another measure \u00e2\u0080\u0093 type-level accuracy (defined as the fraction of words correctly assigned their majority tag), according to which La ng ua ge M etr ic B K 10 E M B K 10 L B F G S G 10 F EA T S B es t F EA T S M ed ia n E ng lis h 1 1 m 1 4 8 . 3 6 8 . 1 5 6 . 0 7 5 . 5 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 0 . 9 6 6 . 4 4 7 . 8 6 6 . 4 D an is h 1 1 m 1 4 2 . 3 6 6 . 7 4 2 . 6 5 8 . 0 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 2 . 1 6 1 . 2 4 3 . 2 6 0 . 7 D ut ch 1 1 m 1 5 3 . 7 6 7 . 0 5 5 . 1 6 4 . 7 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 6 . 4 6 9 . 0 5 1 . 5 6 7 . 3 Po rtu gu es e 1 1 m 1 5 0 . 8 7 5 . 3 4 3 . 2 7 4 . 8 44 .5 69 .2 6 4 . 1 7 4 . 5 5 6 . 5 7 0 . 1 S pa ni sh 1 1 m 1 \u00e2\u0080\u0093 \u00e2\u0080\u0093 4 0 . 6 7 3 . 2 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 8 . 3 6 8 . 9 5 0 . 0 5 7 . 2 Table 4: Comparison of our method (FEATS) to state-of-the-art methods.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Memory usage in PROBING is high, though SRILM is even larger, so where memory is of concern we recommend using TRIE, if it fits in memory.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Papers that use this method or minor variants thereof include Liang (1986), Li et al.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"This is the form of recursive levels in iDafa constructs.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"both MENE and IdentiFinder used more training data than we did (we used only the official MUC 6 and MUC7 training data).","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"The effect of the pruning threshold t0 is shown in Table 5.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"Ltd., then organization will be more probable.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"SRILM\u2019s compact variant has an incredibly expensive destructor, dwarfing the time it takes to perform translation, and so we also modified Moses to avoiding the destructor by calling exit instead of returning normally.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Thus in a two-hanzi word like lflli?J zhong1guo2 (middle country) 'China' there are two syllables, and at the same time two morphemes.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"gaolgaolxing4xing4 'happily' In the particular form of A-not-A reduplication illustrated in (3a), the first syllable of the verb is copied, and the negative markerbu4 'not' is inserted between the copy and the full verb.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"A minimal requirement for building a Chinese word segmenter is obviously a dictionary; furthermore, as has been argued persuasively by Fung and Wu (1994), one will perform much better at segmenting text by using a dictionary constructed with text of the same genre as the text to be segmented.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Table 4 shows the results.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"This line of work has been motivated by empirical findings that the standard EM-learned unsupervised HMM does not exhibit sufficient word tag sparsity.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"2.1 Overview.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"5 68.1 34.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"The first point we need to address is what type of linguistic object a hanzi repre\u00c2\u00ad sents.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"For the automatic evaluation, we used BLEU, since it is the most established metric in the field.","label":1},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"Systems that generally do worse than others will receive a negative one.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"IRSTLM and BerkeleyLM use this state function (and a limit of N \u22121 words), but it is more strict than necessary, so decoders using these packages will miss some recombination opportunities.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"This formulation of the constraint feature is equivalent to the use of a tagging dictionary extracted from the graph using a threshold T on the posterior distribution of tags for a given word type (Eq.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"The dev and test sets were randomly chosen from the EMEA corpus.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Arabic sentences of up to length 63 would need to be.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"There are still some open issues to be resolved with the format, but it represents a first step.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"For each domain, phrases which contain the same keyword are gathered to build a set of phrases (Step 3).","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"5 We choose these two metrics over the Variation Information measure due to the deficiencies discussed in Gao and Johnson (2008).","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"Unfortunately, modifying the model to account for these kind of dependencies is not at all straightforward.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"Co-occurrences among the particles themselves are subject to further syntactic and lexical constraints relative to the stem.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"For a trigram language model, the partial hypotheses are of the form (e0; e; C; j).","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"Information from a sentence is sometimes insufficient to classify a name correctly.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"While the semantic aspect of radicals is by no means completely predictive, the semantic homogeneity of many classes is quite striking: for example 254 out of the 263 examples (97%) of the INSECT class listed by Wieger (1965, 77376) denote crawling or invertebrate animals; similarly 21 out of the 22 examples (95%) of the GHOST class (page 808) denote ghosts or spirits.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"We used C = 1.0 as the L2 regularization constant in (Eq.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"First, we describe how the caseframes are represented and learned.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"There is a guarantee of no crossing brackets but there is no guarantee that a constituent in the tree has the same children as it had in any of the three original parses.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"One may argue with these efforts on normalization, and ultimately their value should be assessed by assessing their impact on inter-annotator agreement.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Previous Work.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"In future work, we plan to follow-up on this approach and investigate other ways that contextual role knowledge can be used.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"We will show that by giving the first model some global features, MENERGI outperforms Borthwick' s reference resolution classifier.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"This is because different judges focused on different language pairs.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"98 15.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"Phrase tables were extracted from the IN and OUT training corpora (not the dev as was used for instance weighting models), and phrase pairs in the intersection of the IN and OUT phrase tables were used as positive examples, with two alternate definitions of negative examples: The classifier trained using the 2nd definition had higher accuracy on a development set.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"= p(fj je) max \u00c3\u0086;e00 j02Cnfjg np(jjj0; J) p(\u00c3\u0086) p\u00c3\u0086(eje0; e00) Qe00 (e0;C n fjg; j 0 )o: The DP equation is evaluated recursively for each hypothesis (e0; e; C; j).","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"They first collect the NE instance pairs and contexts, just like our method.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"Recent work has made significant progress on unsupervised POS tagging (Me\u00c2\u00b4rialdo, 1994; Smith and Eisner, 2005; Haghighi and Klein, 2006; Johnson,2007; Goldwater and Griffiths, 2007; Gao and John son, 2008; Ravi and Knight, 2009).","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"handled given appropriate models.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"If either case is true, then CFLex reports that the anaphor and candidate might be coreferent.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"This formulation of the constraint feature is equivalent to the use of a tagging dictionary extracted from the graph using a threshold T on the posterior distribution of tags for a given word type (Eq.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"However, we note that the final conditional estimates p(s|t) from a given phrase table maximize the likelihood of joint empirical phrase pair counts over a word-aligned corpus.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"We will describe the evaluation of such clusters in the next subsection.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"For instance, for out-ofdomain English-French, Systran has the best BLEU and manual scores.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"5 \u00e2\u0080\u0098Underspecified Rhetorical Markup Language\u00e2\u0080\u0099 6 This confirms the figure given by (Schauer, Hahn.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"ogy (Koskenniemi 1983; Antworth 1990; Tzoukermann and Liberman 1990; Karttunen, Kaplan, and Zaenen 1992; Sproat 1992); we represent the fact that ir, attaches to nouns by allowing t:-transitions from the final states of all noun entries, to the initial state of the sub-WFST representing f,.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"Corpus Step 1 NE pair instances Step 2 Step 1.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"A few pointed out that adequacy should be broken up into two criteria: (a) are all source words covered?","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"SRILM (Stolcke, 2002) is widely used within academia.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"The Recency KS computes the distance between the candidate and the anaphor relative to its scope.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Queries detect the invalid probability, using the node only if it leads to a longer match.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"We implement two data structures: PROBING, designed for speed, and TRIE, optimized for memory.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Given a key k, it estimates the position If the estimate is exact (A[pivot] = k), then the algorithm terminates succesfully.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"We present a coreference resolver called BABAR that uses contextual role knowledge to evaluate possible antecedents for an anaphor.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"Table 5: Effect of the beam threshold on the number of search errors (147 sentences).","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"Each feature concept is akin to a random variable and its occurrence in the text corresponds to a particular instantiation of that random variable.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"Unique Occurrences and Zone (UNIQ): This group of features indicates whether the word is unique in the whole document.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"We do not attempt to identify the types of relationships that are found.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"4 65.9 48.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"t(i).","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"with the number of exactly matching guess trees.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Adam Pauls provided a pre-release comparison to BerkeleyLM and an initial Java interface.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"The feature-based model replaces the emission distribution with a log-linear model, such that: on the word identity x, features checking whether x contains digits or hyphens, whether the first letter of x is upper case, and suffix features up to length 3.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"For example, even if the contexts surrounding an anaphor and candidate match exactly, they are not coreferent if they have substantially different meanings 9 We would be happy to make our manually annotated test data available to others who also want to evaluate their coreference resolver on the MUC4 or Reuters collections.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Given a set of n sentences, we can compute the sample mean x\ufffd and sample variance s2 of the individual sentence judgements xi: The extend of the confidence interval [x\u2212d, x+df can be computed by d = 1.96 \u00b7\ufffdn (6) Pairwise Comparison: As for the automatic evaluation metric, we want to be able to rank different systems against each other, for which we need assessments of statistical significance on the differences between a pair of systems.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Given n-gram counts {cn}Nn=1, we use Flog2 c1] bits per vocabulary identifier and Flog2 cn] per index into the table of ngrams.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"The equation for sampling a single type-level assignment Ti is given by, 0.2 0 5 10 15 20 25 30 Iteration Figure 2: Graph of the one-to-one accuracy of our full model (+FEATS) under the best hyperparameter setting by iteration (see Section 5).","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Our initial experimentation with the evaluation tool showed that this is often too overwhelming.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"The contextual role knowledge that BABAR uses for coreference resolution is derived from this caseframe data.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"3An English sentence with ambiguous PoS assignment can be trivially represented as a lattice similar to our own, where every pair of consecutive nodes correspond to a word, and every possible PoS assignment for this word is a connecting arc.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Chang of Tsinghua University, Taiwan, R.O.C., for kindly providing us with the name corpora.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"When dropping the top and bottom 2.5% the remaining BLEU scores define the range of the confidence interval.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"(Levinger et al., 1995; Goldberg et al., ; Adler et al., 2008)) will make the parser more robust and suitable for use in more realistic scenarios.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"Past work however, has typically associ n = n P (Ti)P (Wi|Ti) = i=1 1 n K n ated these features with token occurrences, typically in an HMM.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"The statistical systems seem to still lag behind the commercial rule-based competition when translating into morphological rich languages, as demonstrated by the results for English-German and English-French.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"The confidence intervals are computed by bootstrap resampling for BLEU, and by standard significance testing for the manual scores, as described earlier in the paper.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"This stage of label propagation results in a tag distribution ri over labels y, which encodes the proportion of times the middle word of ui E Vf aligns to English words vy tagged with label y: The second stage consists of running traditional label propagation to propagate labels from these peripheral vertices Vf\ufffd to all foreign language vertices in the graph, optimizing the following objective: 5 POS Induction After running label propagation (LP), we compute tag probabilities for foreign word types x by marginalizing the POS tag distributions of foreign trigrams ui = x\u2212 x x+ over the left and right context words: where the qi (i = 1, ... , |Vf|) are the label distributions over the foreign language vertices and \u00b5 and \u03bd are hyperparameters that we discuss in \u00a76.4.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"For displaying and querying the annoated text, we make use of the Annis Linguistic Database developed in our group for a large research effort (\u00e2\u0080\u0098Sonderforschungsbereich\u00e2\u0080\u0099) revolving around 9 2.7 Information structure.","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"The number of top-ranked pairs to retain is chosen to optimize dev-set BLEU score.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"We then discuss how we adapt and generalize a boosting algorithm, AdaBoost, to the problem of named entity classification.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"For this experiment, we compare our model with the uniform tag assignment prior (1TW) with the learned prior (+PRIOR).","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"We empirically control for the effect of our heuristics to make sure our pruning does not undermine the objectives of our joint task.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Among these 32 sets, we found the following pairs of sets which have two or more links.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"The ATB segmentation scheme is one of many alternatives.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"The P (W |T , \u00cf\u0088) term in the lexicon component now decomposes as: n P (W |T , \u00cf\u0088) = n P (Wi|Ti, \u00cf\u0088) i=1 n \u00ef\u00a3\u00ab \u00ef\u00a3\u00b6 tions are not modeled by the standard HMM, which = n \u00ef\u00a3\u00ad n P (v|\u00cf\u0088Ti f )\u00ef\u00a3\u00b8 instead can model token-level frequency.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"The restriction can be expressed in terms of the number of uncovered source sentence positions to the left of the rightmost position m in the coverage set.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"For 2 < n < N, we use a hash table mapping from the n-gram to the probability and backoff3.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"The price shocked her.), or same-kind (e.g., Her health insurance paid for the hospital fees, but the automobile insurance did not cover the repair.).","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"To check for this, we do pairwise bootstrap resampling: Again, we repeatedly sample sets of sentences, this time from both systems, and compare their BLEU scores on these sets.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"To facilitate comparison with previous work, we exhaustively evaluate this grammar and two other parsing models when gold segmentation is assumed (\u00c2\u00a75).","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"41.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"A large number of rules is needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classi- However, we show that the use of data can reduce the requirements for supervision to just 7 simple &quot;seed&quot; rules.","label":1},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"First, the model assumes independence between the first and second hanzi of a double given name.","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"By contrast, when we turn to a comparison of the three encoding schemes it is hard to find any significant differences, and the overall impression is that it makes little or no difference which encoding scheme is used, as long as there is some indication of which words are assigned their linear head instead of their syntactic head by the projective parser.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"Maximizing (7) is thus much faster than a typical MERT run. where co(s, t) are the counts from OUT, as in (6).","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Due to many similarly performing systems, we are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"In an experiment on automatic rhetorical parsing, the RST-annotations and PoS tags were used by (Reitter 2003) as a training corpus for statistical classification with Support Vector Machines.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"For the automatic scoring method BLEU, we can distinguish three quarters of the systems.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"We have modified Moses (Koehn et al., 2007) to keep our state with hypotheses; to conserve memory, phrases do not keep state.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"Final (F): The rest of the sentence is processed monotonically taking account of the already covered positions.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Section 4 draws some conclusions from the present state of the effort.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"MENE (without the help of hand-coded systems) has been shown to be somewhat inferior in performance.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"We experimented with increasingly rich grammars read off of the treebank.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"Trying to integrate constituent ordering and choice of referring expressions, (Chiarcos 2003) developed a numerical model of salience propagation that captures various factors of author\u00e2\u0080\u0099s intentions and of information structure for ordering sentences as well as smaller constituents, and picking appropriate referring expressions.10 Chiarcos used the PCC annotations of co-reference and information structure to compute his numerical models for salience projection across the generated texts.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"This is because our corpus is not annotated, and hence does not distinguish between the various words represented by homographs, such as, which could be \/adv jiangl 'be about to' orInc jiang4 '(military) general'-as in 1j\\xiao3jiang4 'little general.'","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"Most computational models for coreference resolution rely on properties of the anaphor and candidate antecedent, such as lexical matching, grammatical and syntactic features, semantic agreement, and positional information.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"We can give a tree pumping lemma for TAG's by adapting the uvwxy-theorem for CFL's since the tree sets of TAG's have independent and context-free paths.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"In this paper, we make a simplifying assumption of one-tag-per-word.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"The first term in the objective function is the graph smoothness regularizer which encourages the distributions of similar vertices (large wij) to be similar.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"As indicated by bolding, for seven out of eight languages the improvements of the \u201cWith LP\u201d setting are statistically significant with respect to the other models, including the \u201cNo LP\u201d setting.11 Overall, it performs 10.4% better than the hitherto state-of-the-art feature-HMM baseline, and 4.6% better than direct projection, when we macro-average the accuracy over all languages.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"The natural baseline (baseline) outperforms the pure IN system only for EMEA\/EP fren.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Table 9: Dev set results for sentences of length \u00e2\u0089\u00a4 70.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"Since our goal is to perform well under these measures we will similarly treat constituents as the minimal substructures for combination.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"Handling Unknown tokens When handling unknown tokens in a language such as Hebrew various important aspects have to be borne in mind.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"segmentation (Table 2).","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"Each ht is a function that predicts a label (+1 or \u20141) on examples containing a particular feature xt, while abstaining on other examples: The prediction of the strong hypothesis can then be written as We now briefly describe how to choose ht and at at each iteration.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"The authors provided us with a ratio between TPT and SRI under different conditions. aLossy compression with the same weights. bLossy compression with retuned weights. ditions make the value appropriate for estimating repeated run times, such as in parameter tuning.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"07 95.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"The method reported in this paper makes use solely of unigram probabilities, and is therefore a zeroeth-order model: the cost of a particular segmentation is estimated as the sum of the costs of the individual words in the segmentation.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Table 6: Incremental dev set results for the manually annotated grammar (sentences of length \u00e2\u0089\u00a4 70).","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"If no candidate satisfies this condition (which is often the case), then the anaphor is left unresolved.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"A more forceful approach for encoding sparsity is posterior regularization, which constrains the posterior to have a small number of expected tag assignments (Grac\u00c2\u00b8a et al., 2009).","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"Then, the German infinitive 'besuchen' and the negation particle 'nicht' are translated.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"The morphological anal\u00c2\u00adysis itself can be handled using well-known techniques from finite-state morphol 9 The initial estimates are derived from the frequencies in the corpus of the strings of hanzi making up.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"The reference medicine for Silapo is EPREX\/ERYPO, which contains epoetin alfa.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"Moreover, the Stanford parser achieves the most exact Leaf Ancestor matches and tagging accuracy that is only 0.1% below the Bikel model, which uses pre-tagged input.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"The accuracies for link were 73% and 86% on two evaluated domains.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"Unfortunately, the best completely unsupervised English POS tagger (that does not make use of a tagging dictionary) reaches only 76.1% accuracy (Christodoulopoulos et al., 2010), making its practical usability questionable at best.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"Still, for about good number of sentences, we do have this direct comparison, which allows us to apply the sign test, as described in Section 2.2.","label":0},{"source_text":"Here we present two algorithms.","target_text":"For example, a good classifier would identify Mrs. Frank as a person, Steptoe & Johnson as a company, and Honduras as a location.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"If the parse contains productions from outside our grammar the machine has no direct method for handling them (e.g. the resulting database query may be syntactically malformed).","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"An example is in (i), where the system fails to group t;,f;?\"$?t!: lin2yang2gang3 as a name, because all three hanzi can in principle be separate words (t;,f; lin2 'wood';?\"$ yang2 'ocean'; ?t!; gang3 'harbor').","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"This approach makes the training objective more complex by adding linear constraints proportional to the number of word types, which is rather prohibitive.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"We divide up each test set into blocks of 20 sentences (100 blocks for the in-domain test set, 53 blocks for the out-of-domain test set), check for each block, if one system has a higher BLEU score than the other, and then use the sign test.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"There exist a few robust broad-coverage parsers that produce non-projective dependency structures, notably Tapanainen and J\u00a8arvinen (1997) and Wang and Harper (2004) for English, Foth et al. (2004) for German, and Holan (2004) for Czech.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"As already pointed out in Section 2.4, current theories diverge not only on the number and definition of relations but also on apects of structure, i.e., whether a tree is sufficient as a representational device or general graphs are required (and if so, whether any restrictions can be placed on these graph\u00e2\u0080\u0099s structures \u00e2\u0080\u0094 cf.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"(If fewer than n rules have Precision greater than pin, we 3Note that taking tlie top n most frequent rules already makes the method robut to low count events, hence we do not use smoothing, allowing low-count high-precision features to be chosen on later iterations. keep only those rules which exceed the precision threshold.) pm,n was fixed at 0.95 in all experiments in this paper.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"08 84.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Because of their size, the examples (Figures 2 to 4) appear at the end of the paper.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"In this subsection, we will report the results of the experiment, in terms of the number of words, phrases or clusters.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"This approach needs a phrase as an initial seed and thus the possible relationships to be extracted are naturally limited.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"7).","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"Finally, we incorporate the instance-weighting model into a general linear combination, and learn weights and mixing parameters simultaneously. where c\u03bb(s, t) is a modified count for pair (s, t) in OUT, u(s|t) is a prior distribution, and y is a prior weight.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"The DempsterShafer rule for combining pdfs is: to {C}, meaning that it is 70% sure the correct hypothesis is C. The intersection of these sets is the null set because these beliefs are contradictory.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"In the first stage, we run a single step of label propagation, which transfers the label distributions from the English vertices to the connected foreign language vertices (say, Vf\ufffd) at the periphery of the graph.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Memory mapping also allows the same model to be shared across processes on the same machine.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"Note that these observa sider suffix features, capitalization features, punctuation, and digit features.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"The major problem for all segmentation systems remains the coverage afforded by the dictionary and the lexical rules used to augment the dictionary to deal with unseen words.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"(b) POS tagging accuracy is lowest for maSdar verbal nouns (VBG,VN) and adjectives (e.g., JJ).","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"Mi(c) is a binary function returning t when parser i (from among the k parsers) suggests constituent c should be in the parse.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Section 3 discusses the applications that have been completed with PCC, or are under way, or are planned for the future.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Note that it is in precision that our over\u00c2\u00ad all performance would appear to be poorer than the reported performance of Chang et al., yet based on their published examples, our system appears to be doing better precisionwise.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"Figure 3 Figure 1.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"2.1 Reliable Case Resolutions.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"While building a machine translation system is a serious undertaking, in future we hope to attract more newcomers to the field by keeping the barrier of entry as low as possible.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Following the system devised under the Qing emperor Kang Xi, hanzi have traditionally been classified according to a set of approximately 200 semantic radicals; members of a radical class share a particular structural component, and often also share a common meaning (hence the term 'semantic').","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"We will return to these issues in the discussion section.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"The parser switching oracle is the upper bound on the accuracy that can be achieved on this set in the parser switching framework.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"In the \u00e2\u0080\u009cPerson \u00e2\u0080\u0093 Person\u00e2\u0080\u009d domain, 618 keywords are found, and in the \u00e2\u0080\u009cCountry \u00e2\u0080\u0093 Country\u00e2\u0080\u009d domain, 303 keywords are found.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"These sequence models-based approaches commonly treat token-level tag assignment as the primary latent variable.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"Zt can be written as follows Following the derivation of Schapire and Singer, providing that W+ > W_, Equ.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"For the automatic scoring method BLEU, we can distinguish three quarters of the systems.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Unfortunately, we have much less data to work with than with the automatic scores.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"7 Acknowledgements.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"Annotators suggested that long sentences are almost impossible to judge.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"For example, the independence assumptions mean that the model fails to capture the dependence between specific and more general features (for example the fact that the feature full.-string=New_York is always seen with the features contains (New) and The baseline method tags all entities as the most frequent class type (organization). contains (York) and is never seen with a feature such as contains (Group) ).","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"The average individual parser accuracy was reduced by more than 5% when we added this new parser, but the precision of the constituent voting technique was the only result that decreased significantly.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"Of these cases, 38 were temporal expressions (either a day of the week or month of the year).","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"Most of these groups follow a phrase-based statistical approach to machine translation.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"The 1-bit sign is almost always negative and the 8-bit exponent is not fully used on the range of values, so in practice this corresponds to quantization ranging from 17 to 20 total bits.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"Parameter Component As in the standard Bayesian HMM (Goldwater and Griffiths, 2007), all distributions are independently drawn from symmetric Dirichlet distributions: 2 Note that t and w denote tag and word sequences respectively, rather than individual tokens or tags.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"We consider properties of the tree sets generated by CFG's, Tree Adjoining Grammars (TAG's), Head Grammars (HG's), Categorial Grammars (CG's), and IG's.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"The monotone search performs worst in terms of both error rates mWER and SSER.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"\"c' 0 + 0 \"0 ' \u00e2\u0080\u00a2 + a n t i g r e e d y x g r e e d y < > c u r r e n t m e t h o d o d i e t . o n l y \u00e2\u0080\u00a2 Taiwan 0 \u00c2\u00b7;; 0 c CD E i5 0\"' 9 9 \u00e2\u0080\u00a2 Mainland \u00e2\u0080\u00a2 \u00e2\u0080\u00a2 \u00e2\u0080\u00a2 \u00e2\u0080\u00a2 -0.30.20.1 0.0 0.1 0.2 Dimension 1 (62%) Figure 7 Classical metric multidimensional scaling of distance matrix, showing the two most significant dimensions.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"Ex: The brigade, which attacked ...","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"The verbal reading arises when the maSdar has an NP argument which, in vocalized text, is marked in the accusative case.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"On several languages, we report performance exceeding that of more complex state-of-the art systems.1","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"In this case we are interested in finding' the maximum probability parse, ri, and Mi is the set of relevant (binary) parsing decisions made by parser i. ri is a parse selected from among the outputs of the individual parsers.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"Depending on the threshold t0, the search algorithm may miss the globally optimal path which typically results in additional translation errors.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"That is, we can use the discourse parser on PCC texts, emulating for instance a \u00e2\u0080\u009cco-reference oracle\u00e2\u0080\u009d that adds the information from our co-reference annotations.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Both of these analyses are shown in Figure 4; fortunately, the correct analysis is also the one with the lowest cost, so it is this analysis that is chosen.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Another option is the closedsource data structures from Sheffield (Guthrie and Hepple, 2010).","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"These make left-to-right query patterns convenient, as the application need only provide a state and the word to append, then use the returned state to append another word, etc.","label":0},{"source_text":"Here we present two algorithms.","target_text":"Each learner is free to pick the labels for these instances.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"1 is given in Fig.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"splitIN captures the verb\/preposition idioms that are widespread in Arabic.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"Chunking is not enough to find such relationships.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"The distribution specifies the relative weight, or importance, of each example \u2014 typically, the weak learner will attempt to minimize the weighted error on the training set, where the distribution specifies the weights.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"splitIN captures the verb\/preposition idioms that are widespread in Arabic.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Step 2.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"In Section 2, we brie y review our approach to statistical machine translation.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"This family represents an attempt to generalize the properties shared by CFG's, HG's, TAG's, and MCTAG's.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"Each round is composed of two stages; each stage updates one of the classifiers while keeping the other classifier fixed.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"On the other hand, when all systems produce muddled output, but one is better, and one is worse, but not completely wrong, a judge is inclined to hand out judgements of 4, 3, and 2.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"(1999) did not report using any dictionaries, but mentioned in a footnote that they have added list membership features, which have helped marginally in certain domains.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"This extends previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and using a simpler training procedure.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"This is consistent with the nature of these two settings: log-linear combination, which effectively takes the intersection of IN and OUT, does relatively better on NIST, where the domains are broader and closer together.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"This withdrawal by the treasury secretary is understandable, though.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"Evaluation results within sets Table 1 shows the evaluation result based on the number of phrases in a set.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"In the interest of testing the robustness of these combining techniques, we added a fourth, simple nonlexicalized PCFG parser.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"Context from the whole document can be important in classifying a named entity.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"There are two key benefits of this model architecture.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"Restrictions We compare our new approach with the word reordering used in the IBM translation approach (Berger et al., 1996).","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"In fact, we found that enabling IRSTLM\u2019s cache made it slightly slower, so results in Table 1 use IRSTLM without caching.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"The human evaluators were a non-native, fluent Arabic speaker (the first author) for the ATB and a native English speaker for the WSJ.7 Table 5 shows type- and token-level error rates for each corpus.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"3 Techniques for automatic vocalization have been studied (Zitouni et al., 2006; Habash and Rambow, 2007).","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"This is not to say that a set of standards by which a particular segmentation would count as correct and another incorrect could not be devised; indeed, such standards have been proposed and include the published PRCNSC (1994) and ROCLING (1993), as well as the unpublished Linguistic Data Consortium standards (ca.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"This is the parse that is closest to the centroid of the observed parses under the similarity metric.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"Several extensions of AdaBoost for multiclass problems have been suggested (Freund and Schapire 97; Schapire and Singer 98).","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"The numbers falling into the location, person, organization categories were 186, 289 and 402 respectively.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"For example, the phrase \u00e2\u0080\u009c's New York-based trust unit,\u00e2\u0080\u009d is not a paraphrase of the other phrases in the \u00e2\u0080\u009cunit\u00e2\u0080\u009d set.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Typical data structures are generalized Bloom filters that guarantee a customizable probability of returning the correct answer.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"This procedure is repeated for T rounds while alternating between the two classifiers.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"NN \u00ef\u00bf\u00bd .e NP NNP NP DTNNP NN \u00ef\u00bf\u00bd .e NP NP NNP NP Table 5: Evaluation of 100 randomly sampled variation nuclei types.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Either save money at any cost - or give priority to education.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"As we shall see, most of the linked sets are paraphrases.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"Our analysis and comparison focuses primarily on the one-to-one accuracy since it is a stricter metric than many-to-one accuracy, but also report many-to-one for completeness.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"Phrase tables were extracted from the IN and OUT training corpora (not the dev as was used for instance weighting models), and phrase pairs in the intersection of the IN and OUT phrase tables were used as positive examples, with two alternate definitions of negative examples: The classifier trained using the 2nd definition had higher accuracy on a development set.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"The dev and test sets were randomly chosen from the EMEA corpus.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"Since the early days of statistical NLP, researchers have observed that a part-of-speech tag distribution exhibits \u00e2\u0080\u009cone tag per discourse\u00e2\u0080\u009d sparsity \u00e2\u0080\u0094 words are likely to select a single predominant tag in a corpus, even when several tags are possible.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"This revealed interesting clues about the properties of automatic and manual scoring.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"In our model, we associate these features at the type-level in the lexicon.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"For instance: if 10 systems participate, and one system does better than 3 others, worse then 2, and is not significant different from the remaining 4, its rank is in the interval 3\u20137.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"However, it is almost universally the case that no clear definition of what constitutes a \"correct\" segmentation is given, so these performance measures are hard to evaluate.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"However, the point of RandLM is to scale to even larger data, compensating for this loss in quality.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Some of these approaches (e.g., Lin, Chiang, and Su [1993]) attempt to identify unknown words, but do not ac\u00c2\u00ad tually tag the words as belonging to one or another class of expression.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"One is the accuracy within a set of phrases which share the same keyword; the other is the accuracy of links.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"E.g. when 'Zahnarzttermin' is aligned to dentist's, the extended lexicon model might learn that 'Zahnarzttermin' actuallyhas to be aligned to both dentist's and ap pointment.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"It may seem surprising to some readers that the interhuman agreement scores reported here are so low.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"\u00e2\u0080\u0094 similar results have been observed across multiple languages.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Because conjunctions are elevated in the parse trees when they separate recursive constituents, we choose the right sister instead of the category of the next word.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"In both cases the investigators were able to achieve significant improvements over the previous best tagging results.","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"We assume that the goal in dependency parsing is to construct a labeled dependency graph of the kind depicted in Figure 1.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"In a few cases, the criteria for correctness are made more explicit.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"4.3 Translation Experiments.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Previous Work.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"We create equivalence classes for verb, noun, and adjective POS categories.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"Again, the idea is that having a picture of syntax, co-reference, and sentence-internal information structure at one\u00e2\u0080\u0099s disposal should aid in finding models of discourse structure that are more explanatory and can be empirically supported.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"If either case is true, then CFLex reports that the anaphor and candidate might be coreferent.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"Because of this threshold, very few NE instance pairs could be used and hence the variety of phrases was also limited.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"We can obtain a letter equivalent CFL defined by a CFG in which the for each rule as above, we have the production A \u2014* A1 Anup where tk (up) = cp.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"Table 4 shows how much the Bayes switching technique uses each of the parsers on the test set.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"Table 1: Syntactic Seeding Heuristics BABAR\u00e2\u0080\u0099s reliable case resolution heuristics produced a substantial set of anaphor\/antecedent resolutions that will be the training data used to learn contextual role knowledge.","label":0},{"source_text":"Here we present two algorithms.","target_text":"Fig.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Assuming unseen objects within each class are equiprobable, their probabilities are given by the Good-Turing theorem as: cis E( n'J.ls) Po oc N * E(N8ls) (2) where p815 is the probability of one unseen hanzi in class cls, E(n'J.15 ) is the expected number of hanzi in cls seen once, N is the total number of hanzi, and E(N(\/ 5 ) is the expected number of unseen hanzi in class cls.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Exposing this information to the decoder will lead to better hypothesis recombination.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"And then there are decisions that systems typically hard-wire, because the linguistic motivation for making them is not well understood yet.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"A detailed description of the search procedure used is given in this patent.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"pronunciation depends upon word affiliation: tfJ is pronounced deO when it is a prenominal modification marker, but di4 in the word \u00c2\u00a7tfJ mu4di4 'goal'; fl; is normally ganl 'dry,' but qian2 in a person's given name.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"To summarize, we provided: The performance of the baseline system is similar to the best submissions in last year\u2019s shared task.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"gaolxing4 'happy' => F.i'JF.i'J Jl!","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"\"c' 0 + 0 \"0 ' \u00e2\u0080\u00a2 + a n t i g r e e d y x g r e e d y < > c u r r e n t m e t h o d o d i e t . o n l y \u00e2\u0080\u00a2 Taiwan 0 \u00c2\u00b7;; 0 c CD E i5 0\"' 9 9 \u00e2\u0080\u00a2 Mainland \u00e2\u0080\u00a2 \u00e2\u0080\u00a2 \u00e2\u0080\u00a2 \u00e2\u0080\u00a2 -0.30.20.1 0.0 0.1 0.2 Dimension 1 (62%) Figure 7 Classical metric multidimensional scaling of distance matrix, showing the two most significant dimensions.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Our open-source (LGPL) implementation is also available for download as a standalone package with minimal (POSIX and g++) dependencies.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"We can do that . IbmS: Yes, wonderful.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"This is demonstrated by average scores over all systems, in terms of BLEU, fluency and adequacy, as displayed in Figure 5.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"We follow the guidelines developed in the TIGER project (Brants et al. 2002) for syntactic annotation of German newspaper text, using the Annotate3 tool for interactive construction of tree structures.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"For \u00c3\u0086 = 1, a new target language word is generated using the trigram language model p(eje0; e00).","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"3.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"For instance: if 10 systems participate, and one system does better than 3 others, worse then 2, and is not significant different from the remaining 4, its rank is in the interval 3\u20137.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Though we do not directly compare state implementations, performance metrics in Table 1 indicate our overall method is faster.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"Unfortunately, Yarowsky's method is not well understood from a theoretical viewpoint: we would like to formalize the notion of redundancy in unlabeled data, and set up the learning task as optimization of some appropriate objective function.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"\u201cin the shadow\u201d.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"For example, from the sentence \u00e2\u0080\u009cMr.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"After tagging a large corpus with an automatic NE tagger, the method tries to find sets of paraphrases automatically without being given a seed phrase or any kinds of cue.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"Performance typically stabilizes across languages after only a few number of iterations.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"In line with perplexity results from Table 1, the PROBING model is the fastest followed by TRIE, and subsequently other packages.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"This larger corpus was kindly provided to us by United Informatics Inc., R.O.C. a set of initial estimates of the word frequencies.9 In this re-estimation procedure only the entries in the base dictionary were used: in other words, derived words not in the base dictionary and personal and foreign names were not used.","label":0},{"source_text":"Here we present two algorithms.","target_text":"So the success of the algorithm may well be due to its success in maximizing the number of unlabeled examples on which the two decision lists agree.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"shows some keywords with their scores.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"The type-level tag assignments T generate features associated with word types W . The tag assignments constrain the HMM emission parameters \u00ce\u00b8.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Since the transducers are built from human-readable descriptions using a lexical toolkit (Sproat 1995), the system is easily maintained and extended.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Rather we believe several methods have to be developed using different heuristics to discover wider variety of paraphrases.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"Each extraction pattern represents a linguistic expression and a syntactic position indicating where a role filler can be found.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"We can make several observations on the cause of errors.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"For example, as Gan (1994) has noted, one can construct examples where the segmen\u00c2\u00ad tation is locally ambiguous but can be determined on the basis of sentential or even discourse context.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"Evaluation results for links","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"0 68.4 48.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"((fii, Q2, Pa) , \u2014\u25a0 (01, i32, 03) , e),(02,e) Oa, en) The path complexity of the tee set generated by a MCTAG is not necessarily context-free.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"rently, some annotations (in particular the connectives and scopes) have already moved beyond the core corpus; the others will grow step by step.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"For the experiments, we use a simple preprocessing step.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"of Articles No.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"\u00ce\u00b2 is the shared hyperparameter for the tag assignment prior and word feature multinomials.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"17 They also provide a set of title-driven rules to identify names when they occur before titles such as $t. 1: xianlshengl 'Mr.' or i:l:itr!J tai2bei3 shi4zhang3 'Taipei Mayor.'","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"Although this property is not structural, it depends on the structural property that sentences can be built from a finite set of clauses of bounded structure as noted by Joshi (1983\/85).","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"The performance of our system on those sentences ap\u00c2\u00ad peared rather better than theirs.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"Other work includes transferring latent topic distributions from source to target language for LM adaptation, (Tam et al., 2007) and adapting features at the sentence level to different categories of sentence (Finch and Sumita, 2008).","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"There are two differences between this method and the DL-CoTrain algorithm: spelling and contextual features, alternating between labeling and learning with the two types of features.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"(2009b) evaluated the Bikel parser using the same ATB split, but only reported dev set results with gold POS tags for sentences of length \u00e2\u0089\u00a4 40.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"As suggested at the end of Section 3, the restrictions that have been specified in the definition of LCFRS's suggest that they can be efficiently recognized.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"Taking \/V, to be the number of examples an algorithm classified correctly (where all gold standard items labeled noise were counted as being incorrect), we calculated two measures of accuracy: See Tab.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"However, this argument is only plausible if the formal framework allows non-projective dependency structures, i.e. structures where a head and its dependents may correspond to a discontinuous constituent.","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"This leads to the best reported performance for robust non-projective parsing of Czech.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"Our full model (\u201cWith LP\u201d) outperforms the unsupervised baselines and the \u201cNo LP\u201d setting for all languages.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"\u00e2\u0080\u0094ested in correlations between prosody and dis course structure.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"This is can not be the only explanation, since the discrepancy still holds, for instance, for out-of-domain French-English, where Systran receives among the best adequacy and fluency scores, but a worse BLEU score than all but one statistical system.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Queries take the form p(wn|wn\u22121 1 ) where wn1 is an n-gram.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"It's not clear how to apply these methods in the unsupervised case, as they required cross-validation techniques: for this reason we use the simpler smoothing method shown here. input to the unsupervised algorithm is an initial, &quot;seed&quot; set of rules.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"The second algorithm builds on a boosting algorithm called AdaBoost (Freund and Schapire 97; Schapire and Singer 98).","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"However, for multinomial models like our LMs and TMs, there is a one to one correspondence between instances and features, eg the correspondence between a phrase pair (s, t) and its conditional multinomial probability p(s1t).","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"Consider the case where IX].","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"In our work, we demonstrate that using a simple na\u00c2\u00a8\u00c4\u00b1veBayes approach also yields substantial performance gains, without the associated training complexity.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"Either save money at any cost - or give priority to education.","label":0},{"source_text":"Here we present two algorithms.","target_text":"(8) can now be rewritten5 as which is of the same form as the function Zt used in AdaBoost.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"One of the strengths of the DempsterShafer model is its natural ability to recognize when several credible hypotheses are still in play.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"Although these authors report better gains than ours, they are with respect to a non-adapted baseline.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Finally, we model the probability of a new transliterated name as the product of PTN and PTN(hanzi;) for each hanzi; in the putative name.13 The foreign name model is implemented as an WFST, which is then summed with the WFST implementing the dictionary, morpho 13 The current model is too simplistic in several respects.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"We will evaluate various specific aspects of the segmentation, as well as the overall segmentation per\u00c2\u00ad formance.","label":1},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"The ATB annotation distinguishes between verbal and nominal readings of maSdar process nominals.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"Then, it can be verified that We can now derive the CoBoost algorithm as a means of minimizing Zco.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"To improve agreement during the revision process, a dual-blind evaluation was performed in which 10% of the data was annotated by independent teams.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"We again assume a training set of n examples {x1 . xri} where the first m examples have labels {y1 ... yin}, and the last (n \u2014 m) examples are unlabeled.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"The state-to-tag mapping is obtained from the best hyperparameter setting for 11 mapping shown in Table 3.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"The development of automatic scoring methods is an open field of research.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Sorted arrays store key-value pairs in an array sorted by key, incurring no space overhead.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"Compared to last year\u2019s shared task, the participants represent more long-term research efforts.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"BABAR applies a DempsterShafer probabilistic model to make resolutions based on evidence from the contextual role knowledge sources as well as general knowledge sources.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"Hence, the different averages of manual scores for the different language pairs reflect the behaviour of the judges, not the quality of the systems on different language pairs.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"(Carlson, Marcu 2001) responded to this situation with relatively precise (and therefore long!)","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"Tsarfaty (2006) used a morphological analyzer (Segal, 2000), a PoS tagger (Bar-Haim et al., 2005), and a general purpose parser (Schmid, 2000) in an integrated framework in which morphological and syntactic components interact to share information, leading to improved performance on the joint task.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"For each extension a new position is added to the coverage set.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"As is described in Section 4, the property of having a derivation tree set that is a local set appears to be useful in showing important properties of the languages generated by the formalisms.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"This style of naming is never required and seems to be losing currency.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"For example, in MUC6, there are four zones (TXT, HL, DATELINE, DD).","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"In (1) the sequencema3lu4 cannot be resolved locally, but depends instead upon broader context; similarly in (2), the sequence :::tcai2neng2 cannot be resolved locally: 1.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"The results clearly indicate increased variation in the ATB relative to the WSJ, but care should be taken in assessing the magnitude of the difference.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"On the other hand, when all systems produce muddled output, but one is better, and one is worse, but not completely wrong, a judge is inclined to hand out judgements of 4, 3, and 2.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"Bikel 75 training trees 5000 10000 15000 Figure 3: Dev set learning curves for sentence lengths \u00e2\u0089\u00a4 70.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Lazy mapping reduces memory requirements by loading pages from disk only as necessary.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"From Thatcher's (1973) work, it is obvious that the complexity of the set of paths from root to frontier of trees in a local set (the tree set of a CFG) is regular'.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Our open-source (LGPL) implementation is also available for download as a standalone package with minimal (POSIX and g++) dependencies.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"We use graph-based label propagation for cross-lingual knowledge transfer and use the projected labels as features in an unsupervised model (Berg- Kirkpatrick et al., 2010).","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"The results in Table 1 show PROBING is 81% faster than TRIE, which is in turn 31% faster than the fastest baseline.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"We performed three experiments to evaluate our techniques.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"3.1 General Knowledge Sources.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"Consider for a set of constituents the isolated constituent precision parser metric, the portion of isolated constituents that are correctly hypothesized.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"This feature imposes constraints Table 1: Features based on the token string that are based on the probability of each name class during training.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"Sometimes extracted phrases by themselves are not meaningful to consider without context, but we set the following criteria.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"If is unique, then a feature (Unique, Zone) is set to 1, where Zone is the document zone where appears.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"By taking the ratio of matching n-grams to the total number of n-grams in the system output, we obtain the precision pn for each n-gram order n. These values for n-gram precision are combined into a BLEU score: The formula for the BLEU metric also includes a brevity penalty for too short output, which is based on the total number of words in the system output c and in the reference r. BLEU is sensitive to tokenization.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"T e r r o r i s m Ca sef ra me Semantic Classes <a ge nt > ass ass ina ted group, human inv esti gat ion int o < N P> event exp lod ed out sid e < N P> building N a t u r a l D i s a s t e r s Ca sef ra me Semantic Classes <a ge nt > inv esti gat ing cau se group, human sur viv or of < N P> event, natphenom hit wit h < N P> attribute, natphenom Figure 3: Semantic Caseframe Expectations Figure 2: Lexical Caseframe Expectations To illustrate how lexical expectations are used, suppose we want to determine whether noun phrase X is the antecedent for noun phrase Y. If they are coreferent, then X and Y should be substitutable for one another in the story.4 Consider these sentences: (S1) Fred was killed by a masked man with a revolver.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"The evaluation framework for the shared task is similar to the one used in last year\u2019s shared task.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"3 58.3 40.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"3.2 Stochastic rhetorical analysis.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"No: Predecessor coverage set Successor coverage set 1 (f1; ;mg n flg ; l0) !","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Statistical methods seem particularly applicable to the problem of unknown-word identification, especially for constructions like names, where the linguistic constraints are minimal, and where one therefore wants to know not only that a particular se\u00c2\u00ad quence of hanzi might be a name, but that it is likely to be a name with some probabil\u00c2\u00ad ity.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Nonetheless, the results of the comparison with human judges demonstrates that there is mileage being gained by incorporating models of these types of words.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"4 7 . 3 8 . 9 2 8 . 8 2 0 . 7 3 2 . 3 3 5 . 2 2 9 . 6 2 7 . 6 1 4 . 2 4 2 . 8 4 5 . 9 4 4 . 3 6 0 . 6 6 1 . 5 4 9 . 9 3 3 . 9 Table 6: Type-level Results: Each cell report the type- level accuracy computed against the most frequent tag of each word type.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"While possible to utilize the feature-based log-linear approach described in Berg-Kirkpatrick et al.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","label":1},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"We used these three parsers to explore parser combination techniques.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"For fair comparison, we have tabulated all results with the size of training data used (Table 5 and Table 6).","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"We call this approach parser switching.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"The precision and recall of similarity switching and constituent voting are both significantly better than the best individual parser, and constituent voting is significantly better than parser switching in precision.4 Constituent voting gives the highest accuracy for parsing the Penn Treebank reported to date.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"For \u00c3\u0086 = 1, a new target language word is generated using the trigram language model p(eje0; e00).","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Aside from adding a simple rule to correct alif deletion caused by the preposition J, no other language-specific processing is performed.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"52 77.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"The input to AdaBoost is a set of training examples ((xi , yi), , (x\u201e.\u201e yrn)).","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"(2010), we adopt a simpler na\u00c2\u00a8\u00c4\u00b1ve Bayes strategy, where all features are emitted independently.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"A high-level relation is agent, which relates an animate nominal to a predicate.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"However, our full model takes advantage of word features not present in Grac\u00c2\u00b8a et al.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"3 61.7 38.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Mandarin exhibits several such processes, including A-not-A question formation, il\u00c2\u00ad lustrated in (3a), and adverbial reduplication, illustrated in (3b): 3.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"The record for wn1 stores the offset at which its extensions begin.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"For example, a story can mention \u00e2\u0080\u009cthe FBI\u00e2\u0080\u009d, \u00e2\u0080\u009cthe White House\u00e2\u0080\u009d, or \u00e2\u0080\u009cthe weather\u00e2\u0080\u009d without any prior referent in the story.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"When the maSdar lacks a determiner, the constituent as a whole resem bles the ubiquitous annexation construct \u00ef\u00bf\u00bd ?f iDafa.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Given names are most commonly two hanzi long, occasionally one hanzi long: there are thus four possible name types, which can be described by a simple set of context-free rewrite rules such as the following: 1.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"As indicated in Figure 1(c), apart from this correct analysis, there is also the analysis taking B ri4 as a word (e.g., a common abbreviation for Japan), along with X:\u00c2\u00a5 wen2zhangl 'essay\/ and f!!.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"For a word whose initCaps might be due to its position rather than its meaning (in headlines, first word of a sentence, etc), the case information of other occurrences might be more accurate than its own.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"The correct ambiguity resolution of the syntactic level therefore helps to resolve the morphological one, and vice versa.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"Roughly speaking, the new algorithm presented in this paper performs a similar search, but instead minimizes a bound on the number of (unlabeled) examples on which two classifiers disagree.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"The distribution specifies the relative weight, or importance, of each example \u2014 typically, the weak learner will attempt to minimize the weighted error on the training set, where the distribution specifies the weights.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"If a candidate has a belief value \u00e2\u0089\u00a5 .50, then we select that candidate as the antecedent for the anaphor.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"For each caseframe, BABAR collects the semantic classes associated with the head nouns of NPs that were extracted by the caseframe.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"To formalize the approach, we introduce four verbgroup states S: Initial (I): A contiguous, initial block of source positions is covered.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"Judges varied in the average score they handed out.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"For example, the Wang, Li, and Chang system fails on the sequence 1:f:p:]nian2 nei4 sa3 in (k) since 1F nian2 is a possible, but rare, family name, which also happens to be written the same as the very common word meaning 'year.'","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"At the very least, we are creating a data resource (the manual annotations) that may the basis of future research in evaluation metrics.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"4.3 Translation Experiments.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"Feature-based HMM Model (Berg- Kirkpatrick et al., 2010): The KM model uses a variety of orthographic features and employs the EM or LBFGS optimization algorithm; Posterior regulariation model (Grac\u00c2\u00b8a et al., 2009): The G10 model uses the posterior regular- ization approach to ensure tag sparsity constraint.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"However, MADA is language-specific and relies on manually constructed dictionaries.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"In the current work morphological analyses and lexical probabilities are derived from a small Treebank, which is by no means the best way to go.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"We often encounter sentences that are highly ambiguous in themselves, without some prior knowledge of the entities concerned.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"The domain is general politics, economics and science.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"Here, $ is the sentence boundary symbol, which is thought to be at position 0 in the target sentence.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"As an indication, in our core corpus, we found an average sentence length of 15.8 words and 1.8 verbs per sentence, whereas a randomly taken sample of ten commentaries from the national papers Su\u00c2\u00a8ddeutsche Zeitung and Frankfurter Allgemeine has 19.6 words and 2.1 verbs per sentence.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"4.4 Chinese Personal Names.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"ogy (Koskenniemi 1983; Antworth 1990; Tzoukermann and Liberman 1990; Karttunen, Kaplan, and Zaenen 1992; Sproat 1992); we represent the fact that ir, attaches to nouns by allowing t:-transitions from the final states of all noun entries, to the initial state of the sub-WFST representing f,.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"If somewhere else in the document we see \u00e2\u0080\u009crestrictions put in place by President Bush\u00e2\u0080\u009d, then we can be surer that Bush is a name.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Following the system devised under the Qing emperor Kang Xi, hanzi have traditionally been classified according to a set of approximately 200 semantic radicals; members of a radical class share a particular structural component, and often also share a common meaning (hence the term 'semantic').","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"This differs from other implementations (Stolcke, 2002; Pauls and Klein, 2011) that use hash tables as nodes in a trie, as explained in the next section.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"The paper is organized as follows: Section 2 explains the different layers of annotation that have been produced or are being produced.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"The wounds are still healing.), entity-attribute (e.g., She 2001), who determined that in their corpus of German computer tests, 38% of relations were lexically signalled.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"We have used the Java-based opennlp maximum entropy package1.","label":0},{"source_text":"Here we present two algorithms.","target_text":"Many statistical or machine-learning approaches for natural language problems require a relatively large amount of supervision, in the form of labeled training examples.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"But for eign learners are often surprised by the verbless predications that are frequently used in Arabic.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"The difference is due to more careful S-NOM NP NP NP VP VBG :: b NP restoring NP ADJP NN :: b NP NN NP NP ADJP DTJJ ADJP DTJJ NN :: b NP NP NP ADJP ADJP DTJJ J ..i NN :: b NP NP NP ADJP ADJP DTJJ NN _;\u00ef\u00bf\u00bd NP PRP DTJJ DTJJ J ..i _;\u00ef\u00bf\u00bd PRP J ..i NN _;\u00ef\u00bf\u00bd NP PRP DTJJ NN _;\u00ef\u00bf\u00bd NP PRP DTJJ J ..i role its constructive effective (b) Stanford (c) Berkeley (d) Bik el (a) Reference Figure 4: The constituent Restoring of its constructive and effective role parsed by the three different models (gold segmentation).","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"For the experiments, we used four newswire corpora, the Los Angeles Times\/Washington Post, The New York Times, Reuters and the Wall Street Journal, all published in 1995.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"7 www.cis.upenn.edu\/\u00e2\u0088\u00bcpdtb\/ 8 www.eml-research.de\/english\/Research\/NLP\/ Downloads had to buy a new car.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"Therefore we opted to use the much 7 The DempsterShafer theory assumes that one of the hypotheses in \u00ce\u00b8 is correct, so eliminating all of the hypotheses violates this assumption.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"We have not to date explored these various options.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"2 60.7 56.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"This essentially means that their model does not result in a true joint inference, as executions for different tasks involve tuning a parameter separately.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"Such discrepancies can be aligned via an intermediate level of PoS tags.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"RandLM is the clear winner in RAM utilization, but is also slower and lower quality.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"IRSTLM\u2019s quantized variant is the inspiration for our quantized variant.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"For the manual scoring, we can distinguish only half of the systems, both in terms of fluency and adequacy.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"It is unique, agrees with the maximum-likelihood distribution, and has the exponential form (Della Pietra et al., 1997): where refers to the outcome, the history (or context), and is a normalization function.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"10 Chinese speakers may object to this form, since the suffix f, menD (PL) is usually restricted to.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"They considered a semi-supervised POS tagging scenario and showed that one can use a graph over trigram types, and edge weights based on distributional similarity, to improve a supervised conditional random field tagger.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Most similar is scrolling queries, wherein left-to-right queries that add one word at a time are optimized.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"The unlabeled data gives many such &quot;hints&quot; that two features should predict the same label, and these hints turn out to be surprisingly useful when building a classifier.","label":0},{"source_text":"Here we present two algorithms.","target_text":"The algorithm builds two classifiers in parallel from labeled and unlabeled data.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Systems that generally do better than others will receive a positive average normalizedjudgement per sentence.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Even with vocalization, there are linguistic categories that are difficult to identify without semantic clues.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"On each language we investigate the contribution of each component of our model.","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"This is in contrast to dependency treebanks, e.g.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"Linear weights are difficult to incorporate into the standard MERT procedure because they are \u201chidden\u201d within a top-level probability that represents the linear combination.1 Following previous work (Foster and Kuhn, 2007), we circumvent this problem by choosing weights to optimize corpus loglikelihood, which is roughly speaking the training criterion used by the LM and TM themselves.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"Search hypotheses are processed separately according to their coverage set C. The best scored hypothesis for each coverage set is computed: QBeam(C) = max e;e0 ;S;j Qe0 (e; S; C; j) The hypothesis (e0; e; S; C; j) is pruned if: Qe0 (e; S; C; j) < t0 QBeam(C); where t0 is a threshold to control the number of surviving hypotheses.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"This material is based upon work supported by the National Science Foundation Graduate Research Fellowship under Grant No.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"The work reported here is closely related to [Ha- segawa et al. 04].","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Note also that the costs currently used in the system are actually string costs, rather than word costs.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"A simple extension will be used to handle this problem.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"and f,.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"The reason why we did not train with both MUC6 and MUC7 training data at the same time is because the task specifications for the two tasks are not identical.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"In this shared task, we were also confronted with this problem, and since we had no funding for paying human judgements, we asked participants in the evaluation to share the burden.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"To be a member of LCFRS a formalism must satisfy two restrictions.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"Hereafter, each pair of NE categories will be called a domain; e.g. the \u00e2\u0080\u009cCompany \u00e2\u0080\u0093 Company\u00e2\u0080\u009d domain, which we will call CC- domain (Step 2).","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"(2009).","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"Step 1.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"0.2 0.1 0.0 -0.1 25 26 27 28 29 30 31 32 -0.2 -0.3 \u2022systran \u2022 ntt 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 20 21 22 23 24 25 26 Fluency Fluency \u2022systran \u2022nrc rali 25 26 27 28 29 30 31 32 0.2 0.1 0.0 -0.1 -0.2 -0.3 cme p \ufffd 20 21 22 23 24 25 26 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 Figure 14: Correlation between manual and automatic scores for English-French 119 In Domain Out of Domain \u2022upv Adequacy -0.9 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 \u2022upv 23 24 25 26 27 28 29 30 31 32 \u2022upc-mr \u2022utd \u2022upc-jmc \u2022uedin-birch \u2022ntt \u2022rali \u2022uedin-birch 16 17 18 19 20 21 22 23 24 25 26 27 Adequacy \u2022upc-mr 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 -0.9 -1.0 -1.1 English-Spanish Fluency \u2022ntt \u2022nrc \u2022rali \u2022uedin-birch -0.2 -0.3 -0.5 \u2022upv 16 17 18 19 20 21 22 23 24 25 26 27 -0.4 nr \u2022 rali Fluency -0.4 \u2022upc-mr utd \u2022upc-jmc -0.5 -0.6 \u2022upv 23 24 25 26 27 28 29 30 31 32 0.2 0.1 -0.0 -0.1 -0.2 -0.3 0.3 0.2 0.1 -0.0 -0.1 -0.6 -0.7 Figure 15: Correlation between manual and automatic scores for English-Spanish 120 English-German In Domain Out of Domain Adequacy Adequacy 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 \u2022upv -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 -0.9 \u2022upv 0.5 0.4 \u2022systran \u2022upc-mr \u2022 \u2022rali 0.3 \u2022ntt 0.2 0.1 -0.0 -0.1 \u2022systran \u2022upc-mr -0.9 9 10 11 12 13 14 15 16 17 18 19 6 7 8 9 10 11 Fluency 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 \u2022upv -0.5 \u2022upv \u2022systran \u2022upc-mr \u2022 Fluency 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 \u2022systran \u2022ntt","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"We have not explored this strategy.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"For locations, tokens are matched against unigrams, and sequences of two consecutive tokens are matched against bigrams.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"Unfortunately, the best completely unsupervised English POS tagger (that does not make use of a tagging dictionary) reaches only 76.1% accuracy (Christodoulopoulos et al., 2010), making its practical usability questionable at best.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"However, their inverted variant implements a reverse trie using less CPU and the same amount of memory7.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"In general, the neighborhoods can be more diverse and we allow a soft label distribution over the vertices.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Section 4 draws some conclusions from the present state of the effort.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"Arabic is a morphologically rich language with a root-and-pattern system similar to other Semitic languages.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"However, dynamic programming can be used to find the shortest tour in exponential time, namely in O(n22n), using the algorithm by Held and Karp.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"4.5 Transliterations of Foreign Words.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Note that Chang, Chen, and Chen (1991), in addition to word-frequency information, include a constraint-satisfication model, so their method is really a hybrid approach.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"IRSTLM and BerkeleyLM use this state function (and a limit of N \u22121 words), but it is more strict than necessary, so decoders using these packages will miss some recombination opportunities.","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"The most informative scheme, Head+Path, gives the highest scores, although with respect to Head the difference is not statistically significant, while the least informative scheme, Path \u2013 with almost the same performance on treebank transformation \u2013 is significantly lower (p < 0.01).","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"The resulting algorithm has a complexity of O(n!).","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"Figure 2: Order in which source positions are visited for the example given in Fig.1.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"This is can not be the only explanation, since the discrepancy still holds, for instance, for out-of-domain French-English, where Systran receives among the best adequacy and fluency scores, but a worse BLEU score than all but one statistical system.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Another way to view the judgements is that they are less quality judgements of machine translation systems per se, but rankings of machine translation systems.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"P St = n. \u00ce\u00b2 T VARIABLES \u00cf\u0088 Y W : Word types (W1 ,.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"0.2 0.1 0.0 -0.1 25 26 27 28 29 30 31 32 -0.2 -0.3 \u2022systran \u2022 ntt 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 20 21 22 23 24 25 26 Fluency Fluency \u2022systran \u2022nrc rali 25 26 27 28 29 30 31 32 0.2 0.1 0.0 -0.1 -0.2 -0.3 cme p \ufffd 20 21 22 23 24 25 26 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 Figure 14: Correlation between manual and automatic scores for English-French 119 In Domain Out of Domain \u2022upv Adequacy -0.9 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 \u2022upv 23 24 25 26 27 28 29 30 31 32 \u2022upc-mr \u2022utd \u2022upc-jmc \u2022uedin-birch \u2022ntt \u2022rali \u2022uedin-birch 16 17 18 19 20 21 22 23 24 25 26 27 Adequacy \u2022upc-mr 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 -0.9 -1.0 -1.1 English-Spanish Fluency \u2022ntt \u2022nrc \u2022rali \u2022uedin-birch -0.2 -0.3 -0.5 \u2022upv 16 17 18 19 20 21 22 23 24 25 26 27 -0.4 nr \u2022 rali Fluency -0.4 \u2022upc-mr utd \u2022upc-jmc -0.5 -0.6 \u2022upv 23 24 25 26 27 28 29 30 31 32 0.2 0.1 -0.0 -0.1 -0.2 -0.3 0.3 0.2 0.1 -0.0 -0.1 -0.6 -0.7 Figure 15: Correlation between manual and automatic scores for English-Spanish 120 English-German In Domain Out of Domain Adequacy Adequacy 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 \u2022upv -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 -0.9 \u2022upv 0.5 0.4 \u2022systran \u2022upc-mr \u2022 \u2022rali 0.3 \u2022ntt 0.2 0.1 -0.0 -0.1 \u2022systran \u2022upc-mr -0.9 9 10 11 12 13 14 15 16 17 18 19 6 7 8 9 10 11 Fluency 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 \u2022upv -0.5 \u2022upv \u2022systran \u2022upc-mr \u2022 Fluency 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 \u2022systran \u2022ntt","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"Almost all annotators expressed their preference to move to a ranking-based evaluation in the future.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"An important aspect of the DempsterShafer model is that it operates on sets of hypotheses.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"In the terrorism domain, 1600 texts were used for training and the 40 test docu X \u00e2\u0088\u00a9Y =\u00e2\u0088\u0085 All sets of hypotheses (and their corresponding belief values) in the current model are crossed with the sets of hypotheses (and belief values) provided by the new evidence.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"In sequential tagging models such as (Adler and Elhadad, 2006; Bar-Haim et al., 2007; Smith et al., 2005) weights are assigned according to a language model The input for the joint task is a sequence W = w1, ... , wn of space-delimited tokens.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"The manual scores are averages over the raw unnormalized scores.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"This WFST represents the segmentation of the text into the words AB and CD, word boundaries being marked by arcs mapping between f and part-of-speech labels.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"Keywords with more than one word In the evaluation, we explained that \u00e2\u0080\u009cchairman\u00e2\u0080\u009d and \u00e2\u0080\u009cvice chairman\u00e2\u0080\u009d are considered paraphrases.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"This procedure is repeated for T rounds while alternating between the two classifiers.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"However, dynamic programming can be used to find the shortest tour in exponential time, namely in O(n22n), using the algorithm by Held and Karp.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"An example of a fairly low-level relation is the affix relation, which holds between a stem morpheme and an affix morpheme, such as f1 -menD (PL).","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"The second experiment is limited to data from PDT.5 The training part of the treebank was projectivized under different encoding schemes and used to train memory-based dependency parsers, which were run on the test part of the treebank, consisting of 7,507 sentences and 125,713 tokens.6 The inverse transformation was applied to the output of the parsers and the result compared to the gold standard test set.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"When translating the sentence monotonically from left to right, the translation of the German finite verb 'kann', which is the left verbal brace in this case, is postponed until the German noun phrase 'mein Kollege' is translated, which is the subject of the sentence.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"7 Conclusion and Future Work.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"The use of ILP in learning the desired grammar significantly increases the computational complexity of this method.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Of course, since the number of attested (phonemic) Mandarin syllables (roughly 1400, including tonal distinctions) is far smaller than the number of morphemes, it follows that a given syllable could in principle be written with any of several different hanzi, depending upon which morpheme is intended: the syllable zhongl could be lfl 'middle,''clock,''end,' or ,'loyal.'","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"The basic strategy is, for a given pair of entity types, to start with some examples, like several famous book title and author pairs; and find expressions which contains those names; then using the found expressions, find more author and book title pairs.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"(2010) reports the best unsupervised results for English.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Word Head Of Complement POS 1 '01 inna \u00e2\u0080\u009cIndeed, truly\u00e2\u0080\u009d VP Noun VBP 2 '01 anna \u00e2\u0080\u009cThat\u00e2\u0080\u009d SBAR Noun IN 3 01 in \u00e2\u0080\u009cIf\u00e2\u0080\u009d SBAR Verb IN 4 01 an \u00e2\u0080\u009cto\u00e2\u0080\u009d SBAR Verb IN Table 1: Diacritized particles and pseudo-verbs that, after orthographic normalization, have the equivalent surface form 0 an.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"\u00e2\u0080\u009d).","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"If two phrases can be used to express the same relationship within an information extraction application (\u00e2\u0080\u009cscenario\u00e2\u0080\u009d), these two phrases are paraphrases.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"It may seem surprising to some readers that the interhuman agreement scores reported here are so low.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"For each caseframe, BABAR collects the semantic classes associated with the head nouns of NPs that were extracted by the caseframe.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Following this method, we repeatedly \u2014 say, 1000 times \u2014 sample sets of sentences from the output of each system, measure their BLEU score, and use these 1000 BLEU scores as basis for estimating a confidence interval.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"It is interesting to note, however, that the ability to produce a bounded number of dependent paths (where two dependent paths can share an unbounded amount of information) does not require machinery as powerful as that used in LFG, FUG and IG's.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"We trained this model by optimizing the following objective function: Note that this involves marginalizing out all possible state configurations z for a sentence x, resulting in a non-convex objective.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"An anti-greedy algorithm, AG: instead of the longest match, take the.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"The sentence length probability p(JjI) is omitted without any loss in performance.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"For each resolution in the training data, BABAR also associates the co-referring expression of an NP with the NP\u00e2\u0080\u0099s caseframe.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"Even if an example like this is not labeled, it can be interpreted as a &quot;hint&quot; that Mr and president imply the same category.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"Yarowsky-cautious does not separate the spelling and contextual features, but does have a limit on the number of rules added at each stage.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"This work was supported in part under the GALE program of the Defense Advanced Research Projects Agency, Contract No.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"The AdaBoost algorithm was developed for supervised learning.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"com t 600 Mountain Avenue, 2c278, Murray Hill, NJ 07974, USA.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"We elected run Moses single-threaded to minimize the impact of RandLM\u2019s cache on memory use.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"The paper is structured as follows.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"We used section 23 as the development set for our combining techniques, and section 22 only for final testing.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"The parser builds dependency graphs by traversing the input from left to right, using a stack to store tokens that are not yet complete with respect to their dependents.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"For speed, we plan to implement the direct-mapped cache from BerkeleyLM.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"Section 4 presents experimen tal results on two corpora: the MUC4 terrorism corpus, and Reuters texts about natural disasters.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"It also incorporates the Good-Turing method (Baayen 1989; Church and Gale 1991) in estimating the likelihoods of previously unseen con\u00c2\u00ad structions, including morphological derivatives and personal names.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"With this restriction the resulting tree sets will have independent paths.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Almost all annotators expressed their preference to move to a ranking-based evaluation in the future.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"With the exception of the Dutch data set, no other processing is performed on the annotated tags.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"For RandLM and IRSTLM, the effect of caching can be seen on speed and memory usage.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Automatic scores are computed on a larger tested than manual scores (3064 sentences vs. 300\u2013400 sentences). collected manual judgements, we do not necessarily have the same sentence judged for both systems (judges evaluate 5 systems out of the 8\u201310 participating systems).","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"(1) CEO of McCann . . .","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"We received submissions from 14 groups from 11 institutions, as listed in Figure 2.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"To lower the barrier of entrance to the competition, we provided a complete baseline MT system, along with data resources.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"Gather phrases using keywords Now, the keyword with the top TF\/ITF score is selected for each phrase.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"The form fmnh, for example, can be understood as the verb \u201clubricated\u201d, the possessed noun \u201cher oil\u201d, the adjective \u201cfat\u201d or the verb \u201cgot fat\u201d.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"When the connective is an adverbial, there is much less clarity as to the range of the spans.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"8 57.3 +F EA TS be st me dia n 50.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"Unfortunately, the best completely unsupervised English POS tagger (that does not make use of a tagging dictionary) reaches only 76.1% accuracy (Christodoulopoulos et al., 2010), making its practical usability questionable at best.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"The perplexity for the trigram language model used is 26:5.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"Nonetheless there is no alternative to Reiche\u00e2\u0080\u0099s plan.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"We resolve this problem by inserting an entry with probability set to an otherwise-invalid value (\u2212oc).","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"No: Predecessor coverage set Successor coverage set 1 (f1; ;mg n flg ; l0) !","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Our code is thread-safe, and integrated into the Moses, cdec, and Joshua translation systems.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Bikel 75 training trees 5000 10000 15000 Figure 3: Dev set learning curves for sentence lengths \u00e2\u0089\u00a4 70.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"7.96 5.55 1 l...................................................................................................................................................................................................J..","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"Xim, } associated with the ith example.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"The 95% confidence intervals for type-level errors are (5580, 9440) for the ATB and (1400, 4610) for the WSJ.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"This has led previous workers to adopt ad hoc linear weighting schemes (Finch and Sumita, 2008; Foster and Kuhn, 2007; L\u00a8u et al., 2007).","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"Smith estimates Lotus will make profit this quarter\u00e2\u0080\u00a6\u00e2\u0080\u009d.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"(Webber et al., 2003)).","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"Second, we show that although the Penn Arabic Treebank is similar to other tree- banks in gross statistical terms, annotation consistency remains problematic.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Maamouri et al.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"Fall.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"While the Bootstrap method is slightly more sensitive, it is very much in line with the sign test on text blocks.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Crucially, the conventional orthographic form of MSA text is unvocalized, a property that results in a deficient graphical representation.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"One may argue with these efforts on normalization, and ultimately their value should be assessed by assessing their impact on inter-annotator agreement.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"All the links in the \u00e2\u0080\u009cCC-domain are shown in Step 4 in subsection 3.2.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"For graph propagation, the hyperparameter v was set to 2 x 10\u22126 and was not tuned.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Examples will usually be accompanied by a translation, plus a morpheme-by-morpheme gloss given in parentheses whenever the translation does not adequately serve this purpose.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Adam Pauls provided a pre-release comparison to BerkeleyLM and an initial Java interface.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"In the first part of the experiment, dependency graphs from the treebanks were projectivized using the algorithm described in section 2.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"In focus is in particular the correlation with rhetorical structure, i.e., the question whether specific rhetorical relations \u00e2\u0080\u0094 or groups of relations in particular configurations \u00e2\u0080\u0094 are signalled by speakers with prosodic means.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"If 7 is an elementary tree, the derivation tree consists of a single node labeled 7.","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"This is appropriate in cases where it is sanctioned by Bayes\u2019 law, such as multiplying LM and TM probabilities, but for adaptation a more suitable framework is often a mixture model in which each event may be generated from some domain.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"conceptual relationship in the discourse.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"G1 and G2 are hanzi, we can estimate the probability of the sequence being a name as the product of: \u00e2\u0080\u00a2 the probability that a word chosen randomly from a text will be a name-p(rule 1), and \u00e2\u0080\u00a2 the probability that the name is of the form 1hanzi-family 2hanzi-given-p(rule 2), and \u00e2\u0080\u00a2 the probability that the family name is the particular hanzi F1-p(rule 6), and \u00e2\u0080\u00a2 the probability that the given name consists of the particular hanzi G1 and G2-p(rule 9) This model is essentially the one proposed in Chang et al.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"The normalized judgement per judge is the raw judgement plus (3 minus average raw judgement for this judge).","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"Using this encoding scheme, the arc from je to Z in Figure 2 would be assigned the label AuxP\u2191Sb (signifying an AuxP that has been lifted from a Sb).","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"We define the lattice L to be the concatenation of the lattices Li corresponding to the input words wi (s.t.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"For locations, tokens are matched against unigrams, and sequences of two consecutive tokens are matched against bigrams.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"In this paper, Section 2 begins by explaining how contextual role knowledge is represented and learned.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Consequently, we cut their evidence values in half to lessen their influence.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"Top 5 Bot to m 5 Go ld NN P NN JJ CD NN S RB S PD T # \u00e2\u0080\u009d , 1T W CD W RB NN S VB N NN PR P$ W DT : MD . +P RI OR CD JJ NN S WP $ NN RR B- , $ \u00e2\u0080\u009d . +F EA TS JJ NN S CD NN P UH , PR P$ # . \u00e2\u0080\u009c Table 5: Type-level English POS Tag Ranking: We list the top 5 and bottom 5 POS tags in the lexicon and the predictions of our models under the best hyperparameter setting.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Unexpectedly, because the ministries of treasury and education both had prepared the teacher plan together.","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"Table 5 shows the overall parsing accuracy attained with the three different encoding schemes, compared to the baseline (no special arc labels) and to training directly on non-projective dependency graphs.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"It is a relatively frequent word in the domain, but it can be used in different extraction scenarios.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"9 www.ling.unipotsdam.de\/sfb\/ Figure 2: Screenshot of Annis Linguistic Database 3.3 Symbolic and knowledge-based.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"(f1; ;mg ; l) 2 (f1; ;mg n fl; l1g ; l0) !","label":0},{"source_text":"Here we present two algorithms.","target_text":"971,746 sentences of New York Times text were parsed using the parser of (Collins 96).1 Word sequences that met the following criteria were then extracted as named entity examples: whose head is a singular noun (tagged NN).","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"We use the universal POS tagset of Petrov et al. (2011) in our experiments.10 This set C consists of the following 12 coarse-grained tags: NOUN (nouns), VERB (verbs), ADJ (adjectives), ADV (adverbs), PRON (pronouns), DET (determiners), ADP (prepositions or postpositions), NUM (numerals), CONJ (conjunctions), PRT (particles), PUNC (punctuation marks) and X (a catch-all for other categories such as abbreviations or foreign words).","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"Because conjunctions are elevated in the parse trees when they separate recursive constituents, we choose the right sister instead of the category of the next word.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"For example, both a chair and a truck would be labeled as artifacts, but this does not at all suggest that they are coreferent.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"With a good hash function, collisions of the full 64bit hash are exceedingly rare: one in 266 billion queries for our baseline model will falsely find a key not present.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"A different notion of information structure, is used in work such as that of (?), who tried to characterize felicitous constituent ordering (theme choice, in particular) that leads to texts presenting information in a natural, \u00e2\u0080\u009cflowing\u00e2\u0080\u009d way rather than with abrupt shifts of attention.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"1 The apparent difficulty of adapting constituency models to non-configurational languages has been one motivation for dependency representations (Hajic\u00cb\u0087 and Zema\u00c2\u00b4nek, 2004; Habash and Roth, 2009).","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"Secondly, some segments in a proposed segment sequence may in fact be seen lexical events, i.e., for some p tag Prf(p \u2014* (s, p)) > 0, while other segments have never been observed as a lexical event before.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Such a classification can be seen as a not-always-correct summary of global features.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"5.1 Parsing Models.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"For the TM, this is: where cI(s, t) is the count in the IN phrase table of pair (s, t), po(s|t) is its probability under the OUT TM, and cI(t) = &quot;s, cI(s', t).","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"This range is collapsed to a number of buckets, typically by taking the hash modulo the number of buckets.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"Confidence Interval: To estimate confidence intervals for the average mean scores for the systems, we use standard significance testing.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"For these models we limit the options provided for OOV words by not considering the entire token as a valid segmentation in case at least some prefix segmentation exists.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"This method, one instance of which we term the \"greedy algorithm\" in our evaluation of our own system in Section 5, involves starting at the beginning (or end) of the sentence, finding the longest word starting (ending) at that point, and then repeating the process starting at the next (previous) hanzi until the end (begin\u00c2\u00ad ning) of the sentence is reached.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"Besides size of training data, the use of dictionaries is another factor that might affect performance.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"This technique was introduced by Clarkson and Rosenfeld (1997) and is also implemented by IRSTLM and BerkeleyLM\u2019s compressed option.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"However, there are phrases which express the same meanings even though they do not share the same keyword.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"While it is possible to derive a closed form solution for this convex objective function, it would require the inversion of a matrix of order |Vf|.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"The first value reports use immediately after loading while the second reports the increase during scoring. dBerkeleyLM is written in Java which requires memory be specified in advance.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"The feature-HMM model works better for all languages, generalizing the results achieved for English by Berg-Kirkpatrick et al. (2010).","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"example, in Northern Mandarin dialects there is a morpheme -r that attaches mostly to nouns, and which is phonologically incorporated into the syllable to which it attaches: thus men2+r (door+R) 'door' is realized as mer2.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":".., Tn ) T W \u00cf\u0086 E w : Token word seqs (obs) t : Token tag assigns (det by T ) PARAMETERS \u00cf\u0088 : Lexicon parameters \u00ce\u00b8 : Token word emission parameters \u00cf\u0086 : Token tag transition parameters \u00cf\u0086 \u00cf\u0086 t1 t2 \u00ce\u00b8 \u00ce\u00b8 w1 w2 K \u00cf\u0086 T tm O K \u00ce\u00b8 E wN m N N Figure 1: Graphical depiction of our model and summary of latent variables and parameters.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Next, we represent the input sentence as an unweighted finite-state acceptor (FSA) I over H. Let us assume the existence of a function Id, which takes as input an FSA A, and produces as output a transducer that maps all and only the strings of symbols accepted by A to themselves (Kaplan and Kay 1994).","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"The baseline simply retains the original labels for all arcs, regardless of whether they have been lifted or not, and the number of distinct labels is therefore simply the number n of distinct dependency types.2 In the first encoding scheme, called Head, we use a new label d\u2191h for each lifted arc, where d is the dependency relation between the syntactic head and the dependent in the non-projective representation, and h is the dependency relation that the syntactic head has to its own head in the underlying structure.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"This variant is tested in the experiments below.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"In this paper, we describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","label":0},{"source_text":"Here we present two algorithms.","target_text":"At first glance, the problem seems quite complex: a large number of rules is needed to cover the domain, suggesting that a large number of labeled examples is required to train an accurate classifier.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"The NIST06 and NIST08 evaluation sets were used for testing.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"The parallel data came from the Europarl corpus (Koehn, 2005) and the ODS United Nations dataset (UN, 2006).","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"For example, in MUC6, there are four zones (TXT, HL, DATELINE, DD).","label":0},{"source_text":"Here we present two algorithms.","target_text":"The task is to learn a function from an input string (proper name) to its type, which we will assume to be one of the categories Person, Organization, or Location.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"The type-level tag assignments T generate features associated with word types W . The tag assignments constrain the HMM emission parameters \u00ce\u00b8.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"SRILM inefficiently stores 64-bit pointers.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"For the manual scoring, we can distinguish only half of the systems, both in terms of fluency and adequacy.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"This decreases the statistical significance of our results compared to those studies.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"Roughly speaking, the new algorithm presented in this paper performs a similar search, but instead minimizes a bound on the number of (unlabeled) examples on which two classifiers disagree.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"The probabilistic version of this procedure is straightforward: We once again assume independence among our various member parsers.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"There are two key benefits of this model architecture.","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"On the other hand, given that all schemes have similar parsing accuracy overall, this means that the Path scheme is the least likely to introduce errors on projective arcs.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"This token may further embed into a larger utterance, e.g., \u2018bcl hneim\u2019 (literally \u201cin-the-shadow the-pleasant\u201d, meaning roughly \u201cin the pleasant shadow\u201d) in which the dominated Noun is modified by a proceeding space-delimited adjective.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"The easiest language pair according to BLEU (English-French: 28.33) received worse manual scores than the hardest (English-German: 14.01).","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"To lower the barrier of entrance to the competition, we provided a complete baseline MT system, along with data resources.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"In this section, we brie y review our translation approach.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"Each ht is a function that predicts a label (+1 or \u20141) on examples containing a particular feature xt, while abstaining on other examples: The prediction of the strong hypothesis can then be written as We now briefly describe how to choose ht and at at each iteration.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"'Malaysia.'","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"We confirm the finding by Callison-Burch et al. (2006) that the rule-based system of Systran is not adequately appreciated by BLEU.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"The idea of distinguishing between general and domain-specific examples is due to Daum\u00b4e and Marcu (2006), who used a maximum-entropy model with latent variables to capture the degree of specificity.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"The following algorithm was then used to induce new rules: Let Count' (x) be the number of times feature x is seen with some known label in the training data.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"In all of our experiments, the binary file (whether mapped or, in the case of most other packages, interpreted) is loaded into the disk cache in advance so that lazy mapping will never fault to disk.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"(Yarowsky 95) describes an algorithm for word-sense disambiguation that exploits redundancy in contextual features, and gives impressive performance.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"Chunking is not enough to find such relationships.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"The computing time is low, since no reordering is carried out.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"Evaluation results within sets Table 1 shows the evaluation result based on the number of phrases in a set.","label":0},{"source_text":"There is no global pruning.","target_text":"In the following, we assume that this word joining has been carried out.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"The particular classifier used depends upon the noun.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"We used a standard one-pass phrase-based system (Koehn et al., 2003), with the following features: relative-frequency TM probabilities in both directions; a 4-gram LM with Kneser-Ney smoothing; word-displacement distortion model; and word count.","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"The Danish Dependency Treebank (DDT) comprises about 100K words of text selected from the Danish PAROLE corpus, with annotation of primary and secondary dependencies (Kromann, 2003).","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"The way judgements are collected, human judges tend to use the scores to rank systems against each other.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"The German finite verbs 'bin' (second example) and 'konnten' (third example) are too far away from the personal pronouns 'ich' and 'Sie' (6 respectively 5 source sentence positions).","label":0},{"source_text":"Here we present two algorithms.","target_text":"Recent results (e.g., (Yarowsky 95; Brill 95; Blum and Mitchell 98)) have suggested that unlabeled data can be used quite profitably in reducing the need for supervision.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"The confidence intervals are computed by bootstrap resampling for BLEU, and by standard significance testing for the manual scores, as described earlier in the paper.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Various segmentation approaches were then compared with human performance: 1.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"Hence we decided to select ten commentaries to form a \u00e2\u0080\u0098core corpus\u00e2\u0080\u0099, for which the entire range of annotation levels was realized, so that experiments with multi-level querying could commence.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"Month Names, Days of the Week, and Numbers: If is initCaps and is one of January, February, . . .","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"As expected, the most informative encoding, Head+Path, gives the highest accuracy with over 99% of all non-projective arcs being recovered correctly in both data sets.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"The normalized judgement per sentence is the raw judgement plus (0 minus average raw judgement for this judge on this sentence).","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"Table 6: Example Translations for the Verbmobil task.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"We can use the preceding linguistic and annotation insights to build a manually annotated Arabic grammar in the manner of Klein and Manning (2003).","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"By contrast, BerkeleyLM\u2019s hash and compressed variants will return incorrect results based on an n \u22121-gram.","label":0},{"source_text":"Here we present two algorithms.","target_text":"The numbers falling into the location, person, organization categories were 186, 289 and 402 respectively.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Each trie node contains a sorted array of entries and they use binary search.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"ni2ya3 and @5:2 xilya3, respectively.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Other strategies could readily 6 As a reviewer has pointed out, it should be made clear that the function for computing the best path is. an instance of the Viterbi algorithm.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"When dropping the top and bottom 2.5% the remaining BLEU scores define the range of the confidence interval.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":" We evaluate the system's performance by comparing its segmentation 'Tudgments\" with the judgments of a pool of human segmenters, and the system is shown to perform quite well.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"2.2 Syntactic structure.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"However, a recent study (Callison-Burch et al., 2006), pointed out that this correlation may not always be strong.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"(Again, the goal of also in structural features.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"2.1 Reliable Case Resolutions.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Given counts cn1 where e.g. c1 is the vocabulary size, total memory consumption, in bits, is Our PROBING data structure places all n-grams of the same order into a single giant hash table.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"(a) I f f fi * fi :1 }'l ij 1\u00c2\u00a7: {1M m m s h e n 3 m e 0 shi2 ho u4 wo 3 cai2 ne ng 2 ke4 fu 2 zh e4 ge 4 ku n4 w h a t ti m e I just be abl e ov er co m e thi s C L dif fic 'When will I be able to overcome this difficulty?'","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"We quantify error categories in both evaluation settings.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"Sequence of Initial Caps (SOIC): In the sentence Even News Broadcasting Corp., noted for its accurate reporting, made the erroneous announcement., a NER may mistake Even News Broadcasting Corp. as an organization name.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"Among the IS-units, the referring expressions are marked as such and will in the second phase receive a label for cognitive status (active, accessible- text, accessible-situation, inferrable, inactive).","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"We briefly review it here for completeness.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"The differences in difficulty are better reflected in the BLEU scores than in the raw un-normalized manual judgements.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"Table 6: Incremental dev set results for the manually annotated grammar (sentences of length \u00e2\u0089\u00a4 70).","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Minimal perfect hashing is used to find the index at which a quantized probability and possibly backoff are stored.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"Across all languages, high performance can be attained by selecting a single tag per word type.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"Given the limited number of judgements we received, we did not try to evaluate this.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"Furthermore, some of the arcs represent lexemes not present in the input tokens (e.g. h\/DT, fl\/POS), however these are parts of valid analyses of the token (cf. super-segmental morphology section 2).","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"There is a sizable literature on Chinese word segmentation: recent reviews include Wang, Su, and Mo (1990) and Wu and Tseng (1993).","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"(b) does the translation have the same meaning, including connotations?","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"There are two possible reasons: (1) the knowledge sources are resolving different cases of anaphora, and (2) the knowledge sources provide multiple pieces of evidence in support of (or against) a candidate, thereby acting synergistically to push the DempsterShafer model over the belief threshold in favor of a single candidate.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"The accuracy is calculated as the ratio of the number of paraphrases to the total number of phrases in the set.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Language models that contain wi must also contain prefixes wi for 1 G i G k. Therefore, when the model is queried for p(wnjwn\u22121 1 ) but the longest matching suffix is wnf , it may return state s(wn1) = wnf since no longer context will be found.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"The improvement is due to the cost of bit-level reads and avoiding reads that may fall in different virtual memory pages.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Table 2 shows BABAR\u00e2\u0080\u0099s performance.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"Our analysis and comparison focuses primarily on the one-to-one accuracy since it is a stricter metric than many-to-one accuracy, but also report many-to-one for completeness.","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"To explore this tradeoff, we have performed experiments with three different encoding schemes (plus a baseline), which are described schematically in Table 1.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"This fact annoyed especially his dog...).","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"Jiang and Zhai (2007) suggest the following derivation, making use of the true OUT distribution po(s, t): where each fi(s, t) is a feature intended to charac- !0\u02c6 = argmax pf(s, t) log p\u03b8(s|t) (8) terize the usefulness of (s, t), weighted by Ai. \u03b8 s,t pf(s, t)po(s, t) log p\u03b8(s|t) The mixing parameters and feature weights (col- != argmax po (s, t) lectively 0) are optimized simultaneously using dev- \u03b8 s,t pf(s, t)co(s, t) log p\u03b8(s|t), set maximum likelihood as before: !\ufffdargmax po (s, t) ! \u03b8 s,t \ufffd\u02c6 = argmax \u02dcp(s, t) log p(s|t; 0).","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"We aligned the texts at a sentence level across all four languages, resulting in 1064 sentence per language.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"(4) is minimized by setting Since a feature may be present in only a few examples, W_ can be in practice very small or even 0, leading to extreme confidence values.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"What is important and is not expressed by the notation is the so-called coverage constraint: each source position j should be 'hit' exactly once by the path of the inverted alignment bI 1 = b1:::bi:::bI . Using the inverted alignments in the maximum approximation, we obtain as search criterion: max I (p(JjI) max eI 1 ( I Yi=1 p(eijei\u00f4\u0080\u0080\u00801 i\u00f4\u0080\u0080\u00802) max bI 1 I Yi=1 [p(bijbi\u00f4\u0080\u0080\u00801; I; J) p(fbi jei)])) = = max I (p(JjI) max eI 1;bI 1 ( I Yi=1 p(eijei\u00f4\u0080\u0080\u00801 i\u00f4\u0080\u0080\u00802) p(bijbi\u00f4\u0080\u0080\u00801; I; J) p(fbi jei)])); where the two products over i have been merged into a single product over i. p(eijei\u00f4\u0080\u0080\u00801 i\u00f4\u0080\u0080\u00802) is the trigram language model probability.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"We also collapse unary chains withidentical basic categories like NP \u00e2\u0086\u0092 NP.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"com t 700 Mountain Avenue, 2d451, Murray Hill, NJ 07974, USA.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"All of these systems were run on data that was not seen during their development.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"gaolxing4 'happy'=> F.i'JF.i'JJI!JI!","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"A novel element of our model is the ability to capture type-level tag frequencies.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"The manual evaluation of scoring translation on a graded scale from 1\u20135 seems to be very hard to perform.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"Since the inclusion of out-ofdomain test data was a very late decision, the participants were not informed of this.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"We train and test on the CoNLL-X training set.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"am 11.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"To check for this, we do pairwise bootstrap resampling: Again, we repeatedly sample sets of sentences, this time from both systems, and compare their BLEU scores on these sets.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"There is twice as much language modelling data, since training data for the machine translation system is filtered against sentences of length larger than 40 words.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"(f1; ;mg n fl1g ; l) 3 (f1; ;mg n fl; l1; l2g ; l0) !","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Cohen and Smith (2007) chose a metric like SParseval (Roark et al., 2006) that first aligns the trees and then penalizes segmentation errors with an edit-distance metric.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"Each model was able to produce hypotheses for all input sentences.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"The reason why we did not train with both MUC6 and MUC7 training data at the same time is because the task specifications for the two tasks are not identical.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"For example, if {N P1, N P2, N P3} are all coreferent, then each NP must be linked to one of the other two NPs.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"\u00e2\u0080\u009d).","label":0},{"source_text":"There is no global pruning.","target_text":"A dynamic programming recursion similar to the one in Eq. 2 is evaluated.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"This leads to a linear combination of domain-specific probabilities, with weights in [0, 1], normalized to sum to 1.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Therefore we opted to use the much 7 The DempsterShafer theory assumes that one of the hypotheses in \u00ce\u00b8 is correct, so eliminating all of the hypotheses violates this assumption.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"TRIE uses less memory and has better locality.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"If two phrases can be used to express the same relationship within an information extraction application (\u00e2\u0080\u009cscenario\u00e2\u0080\u009d), these two phrases are paraphrases.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"Part-of-speech (POS) tag distributions are known to exhibit sparsity \u00e2\u0080\u0094 a word is likely to take a single predominant tag in a corpus.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"If is one of Monday, Tuesday, . . .","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"Table 4 shows translation results for the three approaches.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"mein 5.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"(7), such as the likelihood function used in maximum-entropy problems and other generalized additive models (Lafferty 99).","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"Equ.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"The theory has also been validated empirically.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"Their work used subject-verb, verb-object, and adjective-noun relations to compare the contexts surrounding an anaphor and candidate.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"However, when we pre- tag the input\u00e2\u0080\u0094as is recommended for English\u00e2\u0080\u0094 we notice a 0.57% F1 improvement.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"The problem can be represented as a graph with 2N vertices corresponding to the members of X1 and X2.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"The core of Yarowsky's algorithm is as follows: where h is defined by the formula in equation 2, with counts restricted to training data examples that have been labeled in step 2.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"(Riloff and Jones 99) was brought to our attention as we were preparing the final version of this paper.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"\u2022 Similarly, when the na\u00efve Bayes classifier is configured such that the constituents require estimated probabilities strictly larger than 0.5 to be accepted, there is not enough probability mass remaining on crossing brackets for them to be included in the hypothesis.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"e0; e are the last two target words, C is a coverage set for the already covered source positions and j is the last position visited.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"The task is to learn a function from an input string (proper name) to its type, which we will assume to be one of the categories Person, Organization, or Location.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"9 50.2 +P RI OR be st me dia n 47.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"In this shared task, we were also confronted with this problem, and since we had no funding for paying human judgements, we asked participants in the evaluation to share the burden.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"It is not easy to make a clear definition of \u00e2\u0080\u009cparaphrase\u00e2\u0080\u009d.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"For example, given a sequence F1G1G2, where F1 is a legal single-hanzi family name, and Plural Nouns X g 0 g \"' X X 0 T!i c\"'.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"F1 85 Berkeley 80 Stanford.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"English parsing evaluations usually report results on sentences up to length 40.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Since our destructor is an efficient call to munmap, bypassing the destructor favors only other packages.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"These results are promising and there are several avenues for improving on these results.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"We combined evidence from four contextual role knowledge sources with evidence from seven general knowledge sources using a DempsterShafer probabilistic model.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Our suspicion is that BLEU is very sensitive to jargon, to selecting exactly the right words, and not synonyms that human judges may appreciate as equally good.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"Another question that remains unanswered is to what extent the linguistic information he considers can be handled-or at least approximated-by finite-state language models, and therefore could be directly interfaced with the segmentation model that we have presented in this paper.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"The major problem for our seg\u00c2\u00ad menter, as for all segmenters, remains the problem of unknown words (see Fung and Wu [1994]).","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"The resulting parsers surpass the best previously published performance results for the Penn Treebank.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Hence, there is a total of 29 classes (7 name classes 4 sub-classes 1 not-a-name class).","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"These sequence models-based approaches commonly treat token-level tag assignment as the primary latent variable.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"For humans, this characteristic can impede the acquisition of literacy.","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"For Experiment 1 it is meaningless as a baseline, since it would result in 0% accuracy. mation on path labels but drop the information about the syntactic head of the lifted arc, using the label d\u2191 instead of d\u2191h (AuxP\u2191 instead of AuxP\u2191Sb).","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"In the initial release of the ATB, inter-annotator agreement was inferior to other LDC treebanks (Maamouri et al., 2008).","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"We use a squared loss to penalize neighboring vertices that have different label distributions: kqi \u2212 qjk2 = Ey(qi(y) \u2212 qj(y))2, and additionally regularize the label distributions towards the uniform distribution U over all possible labels Y.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"For each set, the phrases with bracketed frequencies are considered not paraphrases in the set.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"This is less than the 694 judgements 2004 DARPA\/NIST evaluation, or the 532 judgements in the 2005 DARPA\/NIST evaluation.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"pronunciations of individual words; they also need to compute intonational phrase boundaries in long utterances and assign relative prominence to words in those utterances.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"The best answer to this is: many research labs have very competitive systems whose performance is hard to tell apart.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Manual and Automatic Evaluation of Machine Translation between European Languages","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"The current work treats both segmental and super-segmental phenomena, yet we note that there may be more adequate ways to treat supersegmental phenomena assuming Word-Based morphology as we explore in (Tsarfaty and Goldberg, 2008).","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"In all cases, the key is collapsed to its 64-bit hash.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"The reason why we did not train with both MUC6 and MUC7 training data at the same time is because the task specifications for the two tasks are not identical.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"The basic word order is VSO, but SVO, VOS, and VO configurations are also possible.2 Nouns and verbs are created by selecting a consonantal root (usually triliteral or quadriliteral), which bears the semantic core, and adding affixes and diacritics.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"Our results show that BABAR achieves good performance in both domains, and that the contextual role knowledge improves performance, especially on pronouns.","label":1},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"In Equations 1 through 3 we develop the model for constructing our parse using na\u00efve Bayes classification.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Each trie node contains a sorted array of entries and they use binary search.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"None of the models we have presented utilize features associated with a particular constituent (i.e. the label, span, parent label, etc.) to influence parser preference.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"For the translation experiments, Eq. 2 is recursively evaluated.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"Features and context were initially introduced into the models, but they refused to offer any gains in performance.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"In our forth model GTnph we add the definiteness status of constituents following Tsarfaty and Sima\u2019an (2007).","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"computing the precision of the other's judgments relative to this standard.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Email: gale@research.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"Ex: He was found in San Jose, where ...","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"We have not yet tried this.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"Recent results (e.g., (Yarowsky 95; Brill 95; Blum and Mitchell 98)) have suggested that unlabeled data can be used quite profitably in reducing the need for supervision.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"Phrase tables were extracted from the IN and OUT training corpora (not the dev as was used for instance weighting models), and phrase pairs in the intersection of the IN and OUT phrase tables were used as positive examples, with two alternate definitions of negative examples: The classifier trained using the 2nd definition had higher accuracy on a development set.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"While many systems had similar performance, the results offer interesting insights, especially about the relative performance of statistical and rule-based systems.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"The model we use provides a simple framework in which to incorporate a wide variety of lexical information in a uniform way.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"Matsoukas et al (2009) generalize it by learning weights on sentence pairs that are used when estimating relative-frequency phrase-pair probabilities.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"For parsing, the most challenging form of ambiguity occurs at the discourse level.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Therefore, we only score guess\/gold pairs with identical character yields, a condition that allows us to measure parsing, tagging, and segmentation accuracy by ignoring whitespace.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"The number of judgements is additionally fragmented by our breakup of sentences into in-domain and out-of-domain.","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"The details of the transformation procedure are slightly different depending on the encoding schemes: d\u2191h let the linear head be the syntactic head). target arc must have the form wl \u2212\u2192 wm; if no target arc is found, Head is used as backoff. must have the form wl \u2212\u2192 wm and no outgoing arcs of the form wm p'\u2193 \u2212\u2192 wo; no backoff.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"A token that is allCaps will also be initCaps.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"We offer a state function s(wn1) = wn\ufffd where substring wn\ufffd is guaranteed to extend (to the right) in the same way that wn1 does for purposes of language modeling.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"We compare three hash tables: our probing implementation, GCC\u2019s hash set, and Boost\u2019s8 unordered.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"This smooth guarantees that there are no zeroes estimated.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"Again, we can compute average scores for all systems for the different language pairs (Figure 6).","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"Each word is simply tagged with the semantic classes corresponding to all of its senses.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"2 Chinese ?l* han4zi4 'Chinese character'; this is the same word as Japanese kanji..","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"3 Techniques for automatic vocalization have been studied (Zitouni et al., 2006; Habash and Rambow, 2007).","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"From the point of view of recognition, independent paths in the derivation structures suggests that a top-down parser (for example) can work on each branch independently, which may lead to efficient parsing using an algorithm based on the Divide and Conquer technique.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"We focus on this difference between the tree sets of CFG's and IG's, and formalize the notion of dependence between paths in a tree set in Section 3.","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"Still, from a theoretical point of view, projective parsing of non-projective structures has the drawback that it rules out perfect accuracy even as an asymptotic goal.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"The scores and confidence intervals are detailed first in the Figures 7\u201310 in table form (including ranks), and then in graphical form in Figures 11\u201316.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"To evaluate the performance on the segmentation task, we report SEG, the standard harmonic means for segmentation Precision and Recall F1 (as defined in Bar-Haim et al. (2005); Tsarfaty (2006)) as well as the segmentation accuracy SEGTok measure indicating the percentage of input tokens assigned the correct exact segmentation (as reported by Cohen and Smith (2007)).","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"For example Chen and Liu (1992) report precision and recall rates of over 99%, but this counts only the words that occur in the test corpus that also occur in their dictionary.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"This is especially bad with PROBING because it is based on hashing and performs random lookups, but it is not intended to be used in low-memory scenarios.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"Our annotators pointed out that very often they made almost random decisions as to what relation to choose, and where to locate the boundary of a span.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"Previous work that involves the gathering of information from the whole document often uses a secondary classifier, which corrects the mistakes of a primary sentence- based classifier.","label":0},{"source_text":"Here we present two algorithms.","target_text":"(Brin 98) ,describes a system for extracting (author, book-title) pairs from the World Wide Web using an approach that bootstraps from an initial seed set of examples.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"This is motivated by taking \u03b2 po(s|t) to be the parameters of a Dirichlet prior on phrase probabilities, then maximizing posterior estimates p(s|t) given the IN corpus.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"The predominate focus of building systems that translate into English has ignored so far the difficult issues of generating rich morphology which may not be determined solely by local context.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"The first value reports resident size after loading; the second is the gap between post-loading resident memory and peak virtual memory.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"termined by the category of the word that follows it.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"The focus of our work is on the use of contextual role knowledge for coreference resolution.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"It is. based on the traditional character set rather than the simplified character set used in Singapore and Mainland China.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"96 75.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"This is summarized in Equation 5.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"For locations, tokens are matched against unigrams, and sequences of two consecutive tokens are matched against bigrams.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Pairwise comparison is done using the sign test.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"conceptual relationship in the discourse.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"For the automatic scoring method BLEU, we can distinguish three quarters of the systems.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"So, this was a surprise element due to practical reasons, not malice.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"Several such particles may be prefixed onto a single stem, in which case the affixation is subject to strict linear precedence constraints.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"With the exception of the Dutch data set, no other processing is performed on the annotated tags.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"However, we needed to restrict ourselves to these languages in order to be able to evaluate the performance of our approach.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"For example, if CFLex determines that the log- likelihood statistic for the co-occurrence of a particular noun and caseframe corresponds to the 90% confidence level, then CFLex returns .90 as its belief that the anaphor and candidate are coreferent.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"3 60.7 50.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"68 96.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"(2010)\u00e2\u0080\u0099s richest model: optimized via either EM or LBFGS, as their relative performance depends on the language.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"If somewhere else in the document we see \u00e2\u0080\u009crestrictions put in place by President Bush\u00e2\u0080\u009d, then we can be surer that Bush is a name.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Some of these approaches (e.g., Lin, Chiang, and Su [1993]) attempt to identify unknown words, but do not ac\u00c2\u00ad tually tag the words as belonging to one or another class of expression.","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"In practice, we can therefore expect a trade-off such that increasing the amount of information encoded in arc labels will cause an increase in the accuracy of the inverse transformation but a decrease in the accuracy with which the parser can construct the labeled representations.","label":0},{"source_text":"There is no global pruning.","target_text":"Restrictions We compare our new approach with the word reordering used in the IBM translation approach (Berger et al., 1996).","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"They are also labelled for their topicality (yes \/ no), and this annotation is accompanied by a confidence value assigned by the annotator (since it is a more subjective matter).","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"We found that contextual role knowledge was more beneficial for pronouns than for definite noun phrases.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"), and thosethat begin with a verb (\u00ef\u00bf\u00bd ub..i \u00ef\u00bf\u00bdu _..","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"our full model yields 39.3% average error reduction across languages when compared to the basic configuration (1TW).","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"The increase is generally higher for PDT than for DDT, which indicates a greater diversity in non-projective constructions.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Each trie node contains a sorted array of entries and they use binary search.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"If a candidate has a belief value \u00e2\u0089\u00a5 .50, then we select that candidate as the antecedent for the anaphor.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"This can be repeated several times to collect a list of author \/ book title pairs and expressions.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"The second term is a regularizer and encourages all type marginals to be uniform to the extent that is allowed by the first two terms (cf. maximum entropy principle).","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"We thank Felix Hageloh (Hageloh, 2006) for providing us with this version. proposed in (Tsarfaty, 2006).","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Gabbard and Kulick (2008) show that there is significant attachment ambiguity associated with iDafa, which occurs in 84.3% of the trees in our development set.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"The best analysis of the corpus is taken to be the true analysis, the frequencies are re-estimated, and the algorithm is repeated until it converges.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"t(i).","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"The entire treebank is used in the experiment, but only primary dependencies are considered.4 In all experiments, punctuation tokens are included in the data but omitted in evaluation scores.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"72 77.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"In the current work morphological analyses and lexical probabilities are derived from a small Treebank, which is by no means the best way to go.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Table 8a shows that the best model recovers SBAR at only 71.0% F1.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Further, the probing hash table does only one random lookup per query, explaining why it is faster on large data.","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"Although these authors report better gains than ours, they are with respect to a non-adapted baseline.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Furthermore, even the size of the dictionary per se is less important than the appropriateness of the lexicon to a particular test corpus: as Fung and Wu (1994) have shown, one can obtain substantially better segmentation by tailoring the lexicon to the corpus to be segmented.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"We experiment with four values for each hyperparameter resulting in 16 (\u00ce\u00b1, \u00ce\u00b2) combinations: \u00ce\u00b1 \u00ce\u00b2 0.001, 0.01, 0.1, 1.0 0.01, 0.1, 1.0, 10 Iterations In each run, we performed 30 iterations of Gibbs sampling for the type assignment variables W .4 We use the final sample for evaluation.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"However, ince we extracted the test corpus automatically from web sources, the reference translation was not always accurate \u2014 due to sentence alignment errors, or because translators did not adhere to a strict sentence-by-sentence translation (say, using pronouns when referring to entities mentioned in the previous sentence).","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Here, the term frequency (TF) is the frequency of a word in the bag and the inverse term frequency (ITF) is the inverse of the log of the frequency in the entire corpus.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Therefore, performance is more closely tied to the underlying data structure than to the cache.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"5 60.6 Table 3: Multilingual Results: We report token-level one-to-one and many-to-one accuracy on a variety of languages under several experimental settings (Section 5).","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"The translation search is carried out with the category markers and the city names are resubstituted into the target sentence as a postprocessing step.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"The problem is to store these two values for a large and sparse set of n-grams in a way that makes queries efficient.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"The local feature groups are: Non-Contextual Feature: This feature is set to 1 for all tokens.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"We present two algorithms.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"In German, the verbgroup usually consists of a left and a right verbal brace, whereas in English the words of the verbgroup usually form a sequence of consecutive words.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Note that hanzi that are not grouped into dictionary words (and are not identified as single\u00c2\u00ad hanzi words), or into one of the other categories of words discussed in this paper, are left unattached and tagged as unknown words.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"Again, the monotone search performs worst.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"In a model we built with default settings, 1.2% of n + 1-grams were missing their ngram suffix.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"Our second point of comparison is with Grac\u00c2\u00b8a et al.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"This is the first time that we organized a large-scale manual evaluation.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"BABAR then computes statistics over the training examples measuring the frequency with which extraction patterns and noun phrases co-occur in coreference resolutions.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"Cohen and Smith approach this by introducing the \u03b1 hyperparameter, which performs best when optimized independently for each sentence (cf.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"f1; ; Jg denotes a coverage set including all positions from the starting position 1 to position J and j 2 fJ \u00f4\u0080\u0080\u0080L; ; Jg.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"These estimates are in turn combined linearly with relative-frequency estimates from an in-domain phrase table.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"We elected run Moses single-threaded to minimize the impact of RandLM\u2019s cache on memory use.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"12 For English, our Evalb implementation is identical to the most recent reference (EVALB20080701).","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"For displaying and querying the annoated text, we make use of the Annis Linguistic Database developed in our group for a large research effort (\u00e2\u0080\u0098Sonderforschungsbereich\u00e2\u0080\u0099) revolving around 9 2.7 Information structure.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Given that part-of-speech labels are properties of words rather than morphemes, it follows that one cannot do part-of-speech assignment without having access to word-boundary information.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Besides size of training data, the use of dictionaries is another factor that might affect performance.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"01 75.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"This is most severe with RandLM in the multi-threaded case, where each thread keeps a separate cache, exceeding the original model size.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Other good classes include JADE and GOLD; other bad classes are DEATH and RAT.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"However, to classify a token , while Borthwick uses tokens from to (from two tokens before to two tokens after ), we used only the tokens , , and . Even with local features alone, MENERGI outperforms MENE (Borthwick, 1999).","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Given a key k, it estimates the position If the estimate is exact (A[pivot] = k), then the algorithm terminates succesfully.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"While sorted arrays could be used to implement the same data structure as PROBING, effectively making m = 1, we abandoned this implementation because it is slower and larger than a trie implementation.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"The extent to which this constraint is enforced varies greatly across existing methods.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"Future work along these lines will incorporate other layers of annotation, in particular the syntax information.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"In (Reitter, Stede 2003) we went a different way and suggested URML5, an XML format for underspecifying rhetorical structure: a number of relations can be assigned instead of a single one, competing analyses can be represented with shared forests.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"To check for this, we do pairwise bootstrap resampling: Again, we repeatedly sample sets of sentences, this time from both systems, and compare their BLEU scores on these sets.","label":0},{"source_text":"There is no global pruning.","target_text":"Two subjects are each given a calendar and they are asked to schedule a meeting.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"This solution also obviates the need to perform word sense disambiguation.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"We are therefore applying a different method, which has been used at the 2005 DARPA\/NIST evaluation.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"Our initial experimentation with the evaluation tool showed that this is often too overwhelming.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"Adequacy (rank) Fluency (rank) BLEU (rank) upc-jmc (1-7) (1-8) (1-6) lcc (1-6) (1-7) (1-4) utd (1-7) (1-6) (2-7) upc-mr (1-8) (1-6) (1-7) nrc (1-7) (2-6) (8) ntt (1-8) (2-8) (1-7) cmu (3-7) (4-8) (2-7) rali (5-8) (3-9) (3-7) systran (9) (8-9) (10) upv (10) (10) (9) Spanish-English (In Domain) Adequacy (rank) Fluency (rank) BLEU (rank) upc-jmc (1-7) (1-6) (1-5) ntt (1-7) (1-8) (1-5) lcc (1-8) (2-8) (1-4) utd (1-8) (2-7) (1-5) nrc (2-8) (1-9) (6) upc-mr (1-8) (1-6) (7) uedin-birch (1-8) (2-10) (8) rali (3-9) (3-9) (2-5) upc-jg (7-9) (6-9) (9) upv (10) (9-10) (10) German-English (In Domain) Adequacy (rank) Fluency (rank) BLEU (rank) uedin-phi (1-2) (1) (1) lcc (2-7) (2-7) (2) nrc (2-7) (2-6) (5-7) utd (3-7) (2-8) (3-4) ntt (2-9) (2-8) (3-4) upc-mr (3-9) (6-9) (8) rali (4-9) (3-9) (5-7) upc-jmc (2-9) (3-9) (5-7) systran (3-9) (3-9) (10) upv (10) (10) (9) Figure 7: Evaluation of translation to English on in-domain test data 112 English-French (In Domain) Adequacy (rank) Fluency (rank) BLEU (rank) nrc (1-5) (1-5) (1-6) upc-mr (1-4) (1-5) (1-6) upc-jmc (1-6) (1-6) (1-5) systran (2-7) (1-6) (7) utd (3-7) (3-7) (3-6) rali (1-7) (2-7) (1-6) ntt (4-7) (4-7) (1-5) English-Spanish (In Domain) Adequacy (rank) Fluency (rank) BLEU (rank) ms (1-5) (1-7) (7-8) upc-mr (1-4) (1-5) (1-4) utd (1-5) (1-6) (1-4) nrc (2-7) (1-6) (5-6) ntt (3-7) (1-6) (1-4) upc-jmc (2-7) (2-7) (1-4) rali (5-8) (6-8) (5-6) uedin-birch (6-9) (6-10) (7-8) upc-jg (9) (8-10) (9) upv (9-10) (8-10) (10) English-German (In Domain) Adequacy (rank) Fluency (rank) BLEU (rank) upc-mr (1-3) (1-5) (3-5) ntt (1-5) (2-6) (1-3) upc-jmc (1-5) (1-4) (1-3) nrc (2-4) (1-5) (4-5) rali (3-6) (2-6) (1-4) systran (5-6) (3-6) (7) upv (7) (7) (6) Figure 8: Evaluation of translation from English on in-domain test data 113 French-English (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) upc-jmc (1-5) (1-8) (1-4) cmu (1-8) (1-9) (4-7) systran (1-8) (1-7) (9) lcc (1-9) (1-9) (1-5) upc-mr (2-8) (1-7) (1-3) utd (1-9) (1-8) (3-7) ntt (3-9) (1-9) (3-7) nrc (3-8) (3-9) (3-7) rali (4-9) (5-9) (8) upv (10) (10) (10) Spanish-English (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) upc-jmc (1-2) (1-6) (1-3) uedin-birch (1-7) (1-6) (5-8) nrc (2-8) (1-8) (5-7) ntt (2-7) (2-6) (3-4) upc-mr (2-8) (1-7) (5-8) lcc (4-9) (3-7) (1-4) utd (2-9) (2-8) (1-3) upc-jg (4-9) (7-9) (9) rali (4-9) (6-9) (6-8) upv (10) (10) (10) German-English (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) systran (1-4) (1-4) (7-9) uedin-phi (1-6) (1-7) (1) lcc (1-6) (1-7) (2-3) utd (2-7) (2-6) (4-6) ntt (1-9) (1-7) (3-5) nrc (3-8) (2-8) (7-8) upc-mr (4-8) (6-8) (4-6) upc-jmc (4-8) (3-9) (2-5) rali (8-9) (8-9) (8-9) upv (10) (10) (10) Figure 9: Evaluation of translation to English on out-of-domain test data 114 English-French (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) systran (1) (1) (1) upc-jmc (2-5) (2-4) (2-6) upc-mr (2-4) (2-4) (2-6) utd (2-6) (2-6) (7) rali (4-7) (5-7) (2-6) nrc (4-7) (4-7) (2-5) ntt (4-7) (4-7) (3-6) English-Spanish (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) upc-mr (1-3) (1-6) (1-2) ms (1-7) (1-8) (6-7) utd (2-6) (1-7) (3-5) nrc (1-6) (2-7) (3-5) upc-jmc (2-7) (1-6) (3-5) ntt (2-7) (1-7) (1-2) rali (6-8) (4-8) (6-8) uedin-birch (6-10) (5-9) (7-8) upc-jg (8-9) (9-10) (9) upv (9) (8-9) (10) English-German (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) systran (1) (1-2) (1-6) upc-mr (2-3) (1-3) (1-5) upc-jmc (2-3) (3-6) (1-6) rali (4-6) (4-6) (1-6) nrc (4-6) (2-6) (2-6) ntt (4-6) (3-5) (1-6) upv (7) (7) (7) Figure 10: Evaluation of translation from English on out-of-domain test data 115 French-English In domain Out of Domain Adequacy Adequacy 0.3 0.3 \u2022 0.2 0.2 0.1 0.1 -0.0 -0.0 -0.1 -0.1 -0.2 -0.2 -0.3 -0.3 -0.4 -0.4 -0.5 -0.5 -0.6 -0.6 -0.7 -0.7 \u2022upv -0.8 -0.8 21 22 23 24 25 26 27 28 29 30 31 15 16 17 18 19 20 21 22 \u2022upv \u2022systran upcntt \u2022 rali upc-jmc \u2022 cc Fluency 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 \u2022upv -0.5 \u2022systran \u2022upv upc -jmc \u2022 Fluency 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 \u2022 \u2022 \u2022 td t cc upc- \u2022 rali 21 22 23 24 25 26 27 28 29 30 31 15 16 17 18 19 20 21 22 Figure 11: Correlation between manual and automatic scores for French-English 116 Spanish-English Figure 12: Correlation between manual and automatic scores for Spanish-English -0.3 -0.4 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 \u2022upv -0.4 \u2022upv -0.3 In Domain \u2022upc-jg Adequacy 0.3 0.2 0.1 -0.0 -0.1 -0.2 Out of Domain \u2022upc-jmc \u2022nrc \u2022ntt Adequacy upc-jmc \u2022 \u2022 \u2022lcc \u2022 rali \u2022 \u2022rali -0.7 -0.5 -0.6 \u2022upv 23 24 25 26 27 28 29 30 31 32 19 20 21 22 23 24 25 26 27 28 \u2022 \u2022rali Fluency 0.2 0.1 -0.0 -0.1 -0.2 ntt \u2022 upc-mr \u2022lcc \u2022utd \u2022upc-jg \u2022rali Fluency 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 \u2022upc-jmc \u2022 uedin-birch -0.5 -0.5 \u2022upv 23 24 25 26 27 28 29 30 31 32 19 20 21 22 23 24 25 26 27 28 117 In Domain Out of Domain Adequacy Adequacy German-English 15 16 17 18 19 20 21 22 23 24 25 26 27 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 lcc \u2022 upc-jmc \u2022systran \u2022upv Fluency \u2022ula \u2022upc-mr \u2022lcc 15 16 17 18 19 20 21 22 23 24 25 26 27 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 \u2022systran \u2022upv \u2022uedin-phi -jmc \u2022rali \u2022systran -0.3 -0.4 -0.5 -0.6 \u2022upv 12 13 14 15 16 17 18 19 20 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 Fluency uedin-phi \u2022 \u2022 \u2022utd \u2022upc-jmc \u2022upc-mr 0.4 \u2022rali -0.3 -0.4 -0.5 \u2022upv 12 13 14 15 16 17 18 19 20 0.3 0.2 0.1 -0.0 -0.1 -0.2 English-French In Domain Out of Domain Adequacy Adequacy .","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"f;g denotes the empty set, where no source sentence position is covered.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"32 81.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"We call this technique constituent voting.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"The baseline alignment model does not permit that a source word is aligned to two or more target words, e.g. for the translation direction from German toEnglish, the German compound noun 'Zahnarztter min' causes problems, because it must be translated by the two target words dentist's appointment.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"One way to approach this discrepancy is to assume a preceding phase of morphological segmentation for extracting the different lexical items that exist at the token level (as is done, to the best of our knowledge, in all parsing related work on Arabic and its dialects (Chiang et al., 2006)).","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"Using less training data than other systems, our NER is able to perform as well as other state-of-the-art NERs.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"For the English RST-annotated corpus that is made available via LDC, his corresponding result is 62%.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"For.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"This is similar to stacking the different feature instantiations into long (sparse) vectors and computing the cosine similarity between them.","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"For comparison to information-retrieval inspired baselines, eg (L\u00a8u et al., 2007), we select sentences from OUT using language model perplexities from IN.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"This result suggests the benefit of using the automatic discovery method.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"For example, we may have semantic information (e.g. database query operations) associated with the productions in a grammar.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"The ATB annotation guidelines specify that proper nouns should be specified with a flat NP (a).","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"In this paper k = 3 (the three labels are person, organization, location), and we set a = 0.1.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"(2009) also report results on English, but on the reduced 17 tag set, which is not comparable to ours).","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Furthermore, even the size of the dictionary per se is less important than the appropriateness of the lexicon to a particular test corpus: as Fung and Wu (1994) have shown, one can obtain substantially better segmentation by tailoring the lexicon to the corpus to be segmented.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"(Hearst 92) describes a method for extracting hyponyms from a corpus (pairs of words in &quot;isa&quot; relations).","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"The natural baseline approach is to concatenate data from IN and OUT.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"The composition operations are mapped onto operations that use concatenation to define the substrings spanned by the resulting structures.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"1 53.8 47.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Note that the sets of possible classifiers for a given noun can easily be encoded on that noun by grammatical features, which can be referred to by finite-state grammatical rules.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"2.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"Pairwise comparison is done using the sign test.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"The model starts by generating a tag assignment for each word type in a vocabulary, assuming one tag per word.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"If they knew that the first four words in a hypergraph node would never extend to the left and form a 5-gram, then three or even fewer words could be kept in the backward state.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"The overall performance of our joint framework demonstrates that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperforms upper bounds proposed by previous joint disambiguation systems and achieves segmentation and parsing results on a par with state-of-the-art standalone applications results.","label":1},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"For eight judges, ranging k between 1 and 8 corresponded to a precision score range of 90% to 30%, meaning that there were relatively few words (30% of those found by the automatic segmenter) on which all judges agreed, whereas most of the words found by the segmenter were such that one human judge agreed.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"Lattice parsing (Chappelier et al., 1999) is an alternative to a pipeline that prevents cascading errors by placing all segmentation options into the parse chart.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"Tag set As is standard, for all experiments, we set the number of latent model tag states to the size of the annotated tag set.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"The probabilistic version of this procedure is straightforward: We once again assume independence among our various member parsers.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"In figure 4, reverse relations are indicated by `*\u00e2\u0080\u0099 next to the frequency.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"(8) can now be rewritten5 as which is of the same form as the function Zt used in AdaBoost.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"The computing time is low, since no reordering is carried out.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Table 6: Incremental dev set results for the manually annotated grammar (sentences of length \u00e2\u0089\u00a4 70).","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"There have been other kinds of efforts to discover paraphrase automatically from corpora.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"3.2 The DempsterShafer Decision Model.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"The taggers were trained on datasets labeled with the universal tags.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"Cohen and Smith (2007) chose a metric like SParseval (Roark et al., 2006) that first aligns the trees and then penalizes segmentation errors with an edit-distance metric.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"The sign test checks, how likely a sample of better and worse BLEU scores would have been generated by two systems of equal performance.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"It was our hope that this competition, which included the manual and automatic evaluation of statistical systems and one rulebased commercial system, will give further insight into the relation between automatic and manual evaluation.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"21 In Chinese, numerals and demonstratives cannot modify nouns directly, and must be accompanied by.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"If one system is better in 95% of the sample sets, we conclude that its higher BLEU score is statistically significantly better.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"The maximum likelihood estimates (i.e., parameter values which maximize 10) can not be found analytically, but the EM algorithm can be used to hill-climb to a local maximum of the likelihood function from some initial parameter settings.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Section 4 draws some conclusions from the present state of the effort.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"The use of weighted transducers in particular has the attractive property that the model, as it stands, can be straightforwardly interfaced to other modules of a larger speech or natural language system: presumably one does not want to segment Chinese text for its own sake but instead with a larger purpose in mind.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"This feature imposes constraints Table 1: Features based on the token string that are based on the probability of each name class during training.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"Tsarfaty (2006) used a morphological analyzer (Segal, 2000), a PoS tagger (Bar-Haim et al., 2005), and a general purpose parser (Schmid, 2000) in an integrated framework in which morphological and syntactic components interact to share information, leading to improved performance on the joint task.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"Some of the operations will be constant functions, corresponding to elementary structures, and will be written as f () = zi), where each z, is a constant, the string of terminal symbols al an,,,.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"A few pointed out that adequacy should be broken up into two criteria: (a) are all source words covered?","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"There are clearly eight orthographic words in the example given, but if one were doing syntactic analysis one would probably want to consider I'm to consist of two syntactic words, namely I and am.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"The out-of-domain test set differs from the Europarl data in various ways.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"However, reads in the TRIE data structure are more expensive due to bit-level packing, so we found that it is faster to use interpolation search the entire time.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"In the natural disasters domain, agents are often forces of nature, such as hurricanes or wildfires.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"We tokenize MWUs and their POS tags; this reduces the tag set size to 12.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Maamouri et al.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"For instance: if 10 systems participate, and one system does better than 3 others, worse then 2, and is not significant different from the remaining 4, its rank is in the interval 3\u20137.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"The equivalence of CC's with this restriction to TAG's and HG's is, however, still an open problem.","label":0},{"source_text":"Here we present two algorithms.","target_text":"Each ht is a function that predicts a label (+1 or \u20141) on examples containing a particular feature xt, while abstaining on other examples: The prediction of the strong hypothesis can then be written as We now briefly describe how to choose ht and at at each iteration.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"In future work, we plan to follow-up on this approach and investigate other ways that contextual role knowledge can be used.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"If is one of Monday, Tuesday, . . .","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"This section measures performance on shared tasks in order of increasing complexity: sparse lookups, evaluating perplexity of a large file, and translation with Moses.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Since all long sentence translation are somewhat muddled, even a contrastive evaluation between systems was difficult.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"HR0011-06-C-0022.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Interpolation search has a more expensive pivot but performs less pivoting and reads, so it is slow on small data and faster on large data.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"At first glance, we quickly recognize that many systems are scored very similar, both in terms of manual judgement and BLEU.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"The segmenter handles the grouping of hanzi into words and outputs word pronunciations, with default pronunciations for hanzi it cannot group; we focus here primarily on the system's ability to segment text appropriately (rather than on its pronunciation abilities).","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"gaolgaolxing4xing4 'happily' In the particular form of A-not-A reduplication illustrated in (3a), the first syllable of the verb is copied, and the negative markerbu4 'not' is inserted between the copy and the full verb.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"We also mark all tags that dominate a word with the feminine ending :: taa mar buuTa (markFeminine).","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"However, there is a strong relationship between ni1s and the number of hanzi in the class.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"This model admits a simple Gibbs sampling algorithm where the number of latent variables is proportional to the number of word types, rather than the size of a corpus as for a standard HMM sampler (Johnson, 2007).","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"The sentences in the corpus were tagged by a transformation-based chunker and an NE tagger.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Given an anaphor, BABAR identifies the caseframe that would extract it from its sentence.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"While there might be some controversy about the exact definition of such a tagset, these 12 categories cover the most frequent part-of-speech and exist in one form or another in all of the languages that we studied.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"In any event, to date, we have not compared different methods for deriving the set of initial frequency estimates.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"Given a sufficient number of randomly drawn unlabeled examples (i.e., edges), we will induce two completely connected components that together span the entire graph.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"If so, the CF Network reports that the anaphor and candidate may be coreferent.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"In our model, we associate these features at the type-level in the lexicon.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"It is chosen such that the decisions it made in including or excluding constituents are most probable under the models for all of the parsers.","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"In this paper we study the problem of using a parallel corpus from a background domain (OUT) to improve performance on a target domain (IN) for which a smaller amount of parallel training material\u2014though adequate for reasonable performance\u2014is also available.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"There any many techniques for improving language model speed and reducing memory consumption.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"On the first of these-the B set-our system had 64% recall and 86% precision; on the second-the C set-it had 33% recall and 19% precision.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"In this paper we present a stochastic finite-state model for segmenting Chinese text into words, both words found in a (static) lexicon as well as words derived via the above-mentioned productive processes.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"Typically, judges initially spent about 3 minutes per sentence, but then accelerate with experience.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"can expect famous names like Zhou Enlai's to be in many dictionaries, but names such as :fi lf;f; shi2jil-lin2, the name of the second author of this paper, will not be found in any dictionary.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"For the experiments, we used four newswire corpora, the Los Angeles Times\/Washington Post, The New York Times, Reuters and the Wall Street Journal, all published in 1995.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"(Blum and Mitchell 98) describe learning in the following situation: X = X1 X X2 where X1 and X2 correspond to two different &quot;views&quot; of an example.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Finally, this effort is part of a much larger program that we are undertaking to develop stochastic finite-state methods for text analysis with applications to TIS and other areas; in the final section of this paper we will briefly discuss this larger program so as to situate the work discussed here in a broader context.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"From now on all lattice arcs are tagged segments and the assignment of probability P(p \u2014* (s, p)) to lattice arcs proceeds as usual.4 A rather pathological case is when our lexical heuristics prune away all segmentation possibilities and we remain with an empty lattice.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"The problem with these styles of evaluation is that, as we shall demonstrate, even human judges do not agree perfectly on how to segment a given text.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"As a (crude) approximation, we normalize the extraction patterns with respect to active and passive voice and label those extractions as agents or patients.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"As such, global information from the whole context of a document is important to more accurately recognize named entities.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Vocabulary lookup is a sorted array of 64-bit word hashes.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Table 4 Differences in performance between our system and Wang, Li, and Chang (1992).","label":0},{"source_text":"Here we present two algorithms.","target_text":"In addition to the named-entity string (Maury Cooper or Georgia), a contextual predictor was also extracted.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Although this feature helps, we encounter one consequence of variable word order.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"Note that in our formalism a weakhypothesis can abstain.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"In this section, we brie y review our translation approach.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"We observe similar trends when using another measure \u00e2\u0080\u0093 type-level accuracy (defined as the fraction of words correctly assigned their majority tag), according to which La ng ua ge M etr ic B K 10 E M B K 10 L B F G S G 10 F EA T S B es t F EA T S M ed ia n E ng lis h 1 1 m 1 4 8 . 3 6 8 . 1 5 6 . 0 7 5 . 5 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 0 . 9 6 6 . 4 4 7 . 8 6 6 . 4 D an is h 1 1 m 1 4 2 . 3 6 6 . 7 4 2 . 6 5 8 . 0 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 2 . 1 6 1 . 2 4 3 . 2 6 0 . 7 D ut ch 1 1 m 1 5 3 . 7 6 7 . 0 5 5 . 1 6 4 . 7 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 6 . 4 6 9 . 0 5 1 . 5 6 7 . 3 Po rtu gu es e 1 1 m 1 5 0 . 8 7 5 . 3 4 3 . 2 7 4 . 8 44 .5 69 .2 6 4 . 1 7 4 . 5 5 6 . 5 7 0 . 1 S pa ni sh 1 1 m 1 \u00e2\u0080\u0093 \u00e2\u0080\u0093 4 0 . 6 7 3 . 2 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 8 . 3 6 8 . 9 5 0 . 0 5 7 . 2 Table 4: Comparison of our method (FEATS) to state-of-the-art methods.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"They demonstrated this with the comparison of statistical systems against (a) manually post-edited MT output, and (b) a rule-based commercial system.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"We model each parse as the decisions made to create it, and model those decisions as independent events.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"The less favored reading may be selected in certain contexts, however; in the case of , for example, the nominal reading jiang4 will be selected if there is morphological information, such as a following plural affix ir, menD that renders the nominal reading likely, as we shall see in Section 4.3.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Therefore in cases where the segmentation is identical between the two systems we assume that tagging is also identical.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"These sequence models-based approaches commonly treat token-level tag assignment as the primary latent variable.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"pronunciations of individual words; they also need to compute intonational phrase boundaries in long utterances and assign relative prominence to words in those utterances.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"The secondary classifier in (Borthwick, 1999) uses information not just from the current article, but also from the whole test corpus, with an additional feature that indicates if the information comes from the same document or from another document.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"Different sentence structure and rich target language morphology are two reasons for this.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"In a grammar which generates independent paths the derivations of sibling constituents can not share an unbounded amount of information.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"m(S) represents the belief that the correct hypothesis is included in S. The model assumes that evidence also arrives as a probability density function (pdf) over sets of hypotheses.6 Integrating new evidence into the existing model is therefore simply a matter of defining a function to merge pdfs, one representing the current belief system and one representing the beliefs of the new evidence.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"The discovered paraphrases can be a big help to reduce human labor and create a more comprehensive pattern set.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"diesem 3.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"MENE has only been tested on MUC7.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"The Leaf Ancestor metric measures the cost of transforming guess trees to the reference (Sampson and Babarczy, 2003).","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Despite their simplicity, uni- gram weights have been shown as an effective feature in segmentation models (Dyer, 2009).13 The joint parser\/segmenter is compared to a pipeline that uses MADA (v3.0), a state-of-the-art Arabic segmenter, configured to replicate ATB segmentation (Habash and Rambow, 2005).","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"For instance, for out-ofdomain English-French, Systran has the best BLEU and manual scores.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"As the two NE categories are the same, we can\u00e2\u0080\u0099t differentiate phrases with different orders of par ticipants \u00e2\u0080\u0093 whether the buying company or the to-be-bought company comes first.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"Evaluation within a set The evaluation of paraphrases within a set of phrases which share a keyword is illustrated in Figure 4.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"In the current work morphological analyses and lexical probabilities are derived from a small Treebank, which is by no means the best way to go.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"We received submissions from 14 groups from 11 institutions, as listed in Figure 2.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"One conclusion drawn from this annotation effort was that for humans and machines alike, 2 www.sfs.nphil.unituebingen.de\/Elwis\/stts\/ stts.html 3 www.coli.unisb.de\/sfb378\/negra-corpus\/annotate.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"Each feature concept is akin to a random variable and its occurrence in the text corresponds to a particular instantiation of that random variable.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"Other work includes transferring latent topic distributions from source to target language for LM adaptation, (Tam et al., 2007) and adapting features at the sentence level to different categories of sentence (Finch and Sumita, 2008).","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"For the remaining arcs, if the segment is in fact a known lexeme it is tagged as usual, but for the OOV arcs which are valid Hebrew entries lacking tags assignment, we assign all possible tags and then simulate a grammatical constraint.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"For each source word f, the list of its possible translations e is sorted according to p(fje) puni(e), where puni(e) is the unigram probability of the English word e. It is su\u00c3\u0086cient to consider only the best 50 words.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"The way judgements are collected, human judges tend to use the scores to rank systems against each other.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"Annotators argued for the importance of having correct and even multiple references.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"Once HMM parameters (\u00ce\u00b8, \u00cf\u0086) are drawn, a token-level tag and word sequence, (t, w), is generated in the standard HMM fashion: a tag sequence t is generated from \u00cf\u0086.","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"Future work along these lines will incorporate other layers of annotation, in particular the syntax information.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"(1992).","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"This work was supported in part by the Swedish Research Council (621-2002-4207).","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"This is a simple and effective alternative to setting weights discriminatively to maximize a metric such as BLEU.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"For even larger models, storing counts (Talbot and Osborne, 2007; Pauls and Klein, 2011; Guthrie and Hepple, 2010) is a possibility.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"As an indication, in our core corpus, we found an average sentence length of 15.8 words and 1.8 verbs per sentence, whereas a randomly taken sample of ten commentaries from the national papers Su\u00c2\u00a8ddeutsche Zeitung and Frankfurter Allgemeine has 19.6 words and 2.1 verbs per sentence.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"Figure 2: Order in which source positions are visited for the example given in Fig.1.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"As already pointed out in Section 2.4, current theories diverge not only on the number and definition of relations but also on apects of structure, i.e., whether a tree is sufficient as a representational device or general graphs are required (and if so, whether any restrictions can be placed on these graph\u00e2\u0080\u0099s structures \u00e2\u0080\u0094 cf.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"The second concerns the methods used (if any) to ex\u00c2\u00ad tend the lexicon beyond the static list of entries provided by the machine-readable dictionary upon which it is based.","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"In order to facilitate this task, we extend the set of arc labels to encode information about lifting operations.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"The maximum likelihood estimates (i.e., parameter values which maximize 10) can not be found analytically, but the EM algorithm can be used to hill-climb to a local maximum of the likelihood function from some initial parameter settings.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"The zone to which a token belongs is used as a feature.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"Our evaluation includes both weighted and un- weighted lattices.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"3.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Thus we opted not to take the step of creating more precise written annotation guidelines (as (Carlson, Marcu 2001) did for English), which would then allow for measuring inter-annotator agreement.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"If is not initCaps, then (not-initCaps, ) is set to 1.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"First, any grammar must involve a finite number of elementary structures, composed using a finite number of composition operations.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":".","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"There are two weaknesses in Chang et al.'s model, which we improve upon.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"The second weakness is purely conceptual, and probably does not affect the per\u00c2\u00ad formance of the model.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"For all other recursive NPs, we add a common annotation to the POS tag of the head (recursiveNPHead).","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"For this example, since the sequence Even News Broadcasting Corp. only appears once in the document, its longest substring that occurs in the same document is News Broadcasting Corp. In this case, News has an additional feature of I begin set to 1, Broadcasting has an additional feature of I continue set to 1, and Corp. has an additional feature of I end set to 1.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"More complex approaches such as the relaxation technique have been applied to this problem Fan and Tsai (1988}.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"kann 7.nicht 8.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"We present KenLM, a library that implements two data structures for efficient language model queries, reducing both time and costs.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"Fall.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"This is a standard adaptation problem for SMT.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"We evaluate our approach on eight European languages (\u00a76), and show that both our contributions provide consistent and statistically significant improvements.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"This design leads to a significant reduction in the computational complexity of training and inference.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Time includes all queries but excludes random number generation and data structure population.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"We asked six native speakers-three from Taiwan (TlT3), and three from the Mainland (M1M3)-to segment the corpus.","label":0},{"source_text":"Here we present two algorithms.","target_text":"When this feature type was included, CoBoost chose this default feature at an early iteration, thereby giving non-abstaining pseudo-labels for all examples, with eventual convergence to the two classifiers agreeing by assigning the same label to almost all examples.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"Step 3.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"The label for a test example with features x is then defined as In this paper we define h(x, y) as the following function of counts seen in training data: Count(x,y) is the number of times feature x is seen with label y in training data, Count(x) = EyEy Count(x, y). a is a smoothing parameter, and k is the number of possible labels.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"5.1 Parsing Models.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"Our full model (\u201cWith LP\u201d) outperforms the unsupervised baselines and the \u201cNo LP\u201d setting for all languages.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"In principle a feature could be an arbitrary predicate of the (spelling, context) pair; for reasons that will become clear, features are limited to querying either the spelling or context alone.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"For example, if X and Y are coreferent, then both X and Y are considered to co-occur with the caseframe that extracts X as well as the caseframe that extracts Y. We will refer to the set of nouns that co-occur with a caseframe as the lexical expectations of the case- frame.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"The bootstrap method has been critized by Riezler and Maxwell (2005) and Collins et al. (2005), as being too optimistic in deciding for statistical significant difference between systems.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"Unexpectedly, because the ministries of treasury and education both had prepared the teacher plan together.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"In this paper, we have presented a new, e\u00c3\u0086cient DP-based search procedure for statistical machine translation.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"On the first of these-the B set-our system had 64% recall and 86% precision; on the second-the C set-it had 33% recall and 19% precision.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"For example, out of 905 phrases in the CC- domain, 211 phrases contain keywords found in step 2.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"1 2 3.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Keys to the table are hashed, using for example Austin Appleby\u2019s MurmurHash2, to integers evenly distributed over a large range.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Raj and Whittaker (2003) show that integers in a trie implementation can be compressed substantially.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"In a model we built with default settings, 1.2% of n + 1-grams were missing their ngram suffix.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"We have described grammar state splits that significantly improve parsing performance, catalogued parsing errors, and quantified the effect of segmentation errors.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Both of these analyses are shown in Figure 4; fortunately, the correct analysis is also the one with the lowest cost, so it is this analysis that is chosen.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"level.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"Inspection of the data shows that at n = 2500, the two classifiers both give labels on 44,281 (49.2%) of the unlabeled examples, and give the same label on 99.25% of these cases.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"This is demonstrated by average scores over all systems, in terms of BLEU, fluency and adequacy, as displayed in Figure 5.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Finally, we model the probability of a new transliterated name as the product of PTN and PTN(hanzi;) for each hanzi; in the putative name.13 The foreign name model is implemented as an WFST, which is then summed with the WFST implementing the dictionary, morpho 13 The current model is too simplistic in several respects.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"Search hypotheses are processed separately according to their coverage set C. The best scored hypothesis for each coverage set is computed: QBeam(C) = max e;e0 ;S;j Qe0 (e; S; C; j) The hypothesis (e0; e; S; C; j) is pruned if: Qe0 (e; S; C; j) < t0 QBeam(C); where t0 is a threshold to control the number of surviving hypotheses.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"2 62.2 39.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Each word is simply tagged with the semantic classes corresponding to all of its senses.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"A direct-mapped cache makes BerkeleyLM faster on repeated queries, but their fastest (scrolling) cached version is still slower than uncached PROBING, even on cache-friendly queries.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"Phrase tables were extracted from the IN and OUT training corpora (not the dev as was used for instance weighting models), and phrase pairs in the intersection of the IN and OUT phrase tables were used as positive examples, with two alternate definitions of negative examples: The classifier trained using the 2nd definition had higher accuracy on a development set.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"Thus, M works in logspace and recognition can be done on a deterministic TM in polynomial tape.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"First, we evaluated BABAR using only the seven general knowledge sources.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"They contain about 200M words (25M, 110M, 40M and 19M words, respectively).","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"4.4 Chinese Personal Names.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Our code is thread-safe, and integrated into the Moses, cdec, and Joshua translation systems.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Bikel 75 training trees 5000 10000 15000 Figure 3: Dev set learning curves for sentence lengths \u00e2\u0089\u00a4 70.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"The tool we use is MMAX8, which has been specifically designed for marking co-reference.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"We evaluated the results based on two metrics.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"The precision and recall of similarity switching and constituent voting are both significantly better than the best individual parser, and constituent voting is significantly better than parser switching in precision.4 Constituent voting gives the highest accuracy for parsing the Penn Treebank reported to date.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"0 68.4 48.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"While we used the standard metrics of the community, the we way presented translations and prompted for assessment differed from other evaluation campaigns.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"10.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"2.2.3 Lexical Caseframe Expectations The second type of contextual role knowledge learned by BABAR is Lexical Caseframe Expectations, which are used by the CFLex knowledge source.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"We showed that any system defined in this way can be recognized in polynomial time.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"This stage of label propagation results in a tag distribution ri over labels y, which encodes the proportion of times the middle word of ui E Vf aligns to English words vy tagged with label y: The second stage consists of running traditional label propagation to propagate labels from these peripheral vertices Vf\ufffd to all foreign language vertices in the graph, optimizing the following objective: 5 POS Induction After running label propagation (LP), we compute tag probabilities for foreign word types x by marginalizing the POS tag distributions of foreign trigrams ui = x\u2212 x x+ over the left and right context words: where the qi (i = 1, ... , |Vf|) are the label distributions over the foreign language vertices and \u00b5 and \u03bd are hyperparameters that we discuss in \u00a76.4.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":".t The path set of tree sets at level k +1 have the complexity of the string language of level k. The independence of paths in a tree set appears to be an important property.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Non-empty buckets contain an entry belonging to them or to a preceding bucket where a conflict occurred.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"A promising direction for future work is to explicitly model a distribution over tags for each word type.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Linear probing places at most one entry in each bucket.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"The observed performance gains, coupled with the simplicity of model implementation, makes it a compelling alternative to existing more complex counterparts.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Statistics are printed before Moses exits and after parts of the decoder have been destroyed.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"For example, a person\u00e2\u0080\u0099s full name will match with just their last name (e.g., \u00e2\u0080\u009cGeorge Bush\u00e2\u0080\u009d and \u00e2\u0080\u009cBush\u00e2\u0080\u009d), and a company name will match with and without a corporate suffix (e.g., \u00e2\u0080\u009cIBM Corp.\u00e2\u0080\u009d and \u00e2\u0080\u009cIBM\u00e2\u0080\u009d).","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"Much of the study of grammatical systems in computational linguistics has been focused on the weak generative capacity of grammatical formalism.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"This suggests that different types of anaphora may warrant different treatment: definite NP resolution may depend more on lexical semantics, while pronoun resolution may depend more on contextual semantics.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"In fact, it is very difficult to maintain consistent standards, on what (say) an adequacy judgement of 3 means even for a specific language pair.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"Fort= 1,...,T:","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"Moreover, if we want to test on a huge test corpus, indexing the whole corpus might prove computationally expensive.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"From the point of view of computational implementation this can be problematic, since the inclusion of non-projective structures makes the parsing problem more complex and therefore compromises efficiency and in practice also accuracy and robustness.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"As a first step towards modeling transliterated names, we have collected all hanzi occurring more than once in the roughly 750 foreign names in our dictionary, and we estimate the probabil\u00c2\u00ad ity of occurrence of each hanzi in a transliteration (pTN(hanzi;)) using the maximum likelihood estimate.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"The motivation for our more informal approach was the intuition that there are so many open problems in rhetorical analysis (and more so for German than for English; see below) that the main task is qualitative investigation, whereas rigorous quantitative analyses should be performed at a later stage.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"2.3 Assigning Evidence Values.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Previous work deals with this problem by correcting inconsistencies between the named entity classes assigned to different occurrences of the same entity (Borthwick, 1999; Mikheev et al., 1998).","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"Ablation Analysis We evaluate the impact of incorporating various linguistic features into our model in Table 3.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"Finally, we make some improvements to baseline approaches.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"At the same time, the n-gram error rate is sensitive to samples with extreme n-gram counts.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"Because of this, we retokenized and lowercased submitted output with our own tokenizer, which was also used to prepare the training and test data.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"Each decision determines the inclusion or exclusion of a candidate constituent.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"Both parameters depend on a single hyperparameter \u00ce\u00b1.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"For each token , zero, one, or more of the features in each feature group are set to 1.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"The rhetorical structure annotations of PCC have all been converted to URML.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"This is a standard adaptation problem for SMT.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"5 68.1 34.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"We make the assumption that for each example, both xi,. and x2,2 alone are sufficient to determine the label yi.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"The statistical systems seem to still lag behind the commercial rule-based competition when translating into morphological rich languages, as demonstrated by the results for English-German and English-French.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"However, we do not explore this possibility in the current work.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"First, we aim to explicitly characterize examples from OUT as belonging to general language or not.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"Up to now, most IE researchers have been creating paraphrase knowledge (or IE patterns) by hand and for specific tasks.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"In this specific case, as these two titles could fill the same column of an IE table, we regarded them as paraphrases for the evaluation.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"\u00ce\u00b8 has a belief value of 1.0, indicating complete certainty that the correct hypothesis is included in the set, and a plausibility value of 1.0, indicating that there is no evidence for competing hypotheses.5 As evidence is collected and the likely hypotheses are whittled down, belief is redistributed to subsets of \u00ce\u00b8.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"We incorporate instance weighting into a mixture-model framework, and find that it yields consistent improvements over a wide range of baselines.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Due to the dramatic fiscal situation in Brandenburg she now surprisingly withdrew legislation drafted more than a year ago, and suggested to decide on it not before 2003.","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"In the first part of the experiment, dependency graphs from the treebanks were projectivized using the algorithm described in section 2.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"In this case, Maury Cooper is extracted.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"(Riloff and Jones 99) was brought to our attention as we were preparing the final version of this paper.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"For example, we might have VP \u00e2\u0086\u0092 VB NP PP, where the NP is the subject.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Within the RST \u00e2\u0080\u009cuser community\u00e2\u0080\u009d there has also been discussion whether two levels of discourse structure should not be systematically distinguished (intentional versus informational).","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"Previous work deals with this problem by correcting inconsistencies between the named entity classes assigned to different occurrences of the same entity (Borthwick, 1999; Mikheev et al., 1998).","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"7 Acknowledgements.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Base NPs are the other significant category of nominal phrases.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"Now assume we have n pairs (xi,, x2,i) drawn from X1 X X2, where the first m pairs have labels whereas for i = m+ 1...n the pairs are unlabeled.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"We can represent any derived tree of a TAG by the two substrings that appear in its frontier, and then define how the adjunction operation concatenates the substrings.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"The accuracy of the sets in representing paraphrase ranged from 73% to 99%, depending on the NE categories and set sizes; the accuracy of the links for two evaluated domains was 73% and 86%.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"We maintain a separate array for each length n containing all n-gram entries sorted in suffix order.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"Such sequences are given additional features of A begin, A continue, or A end, and the acronym is given a feature A unique.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"Although matching is done at the sentence level, this information is subsequently discarded when all matches are pooled.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"In Table 1 we see with very few exceptions that the isolated constituent precision is less than 0.5 when we use the constituent label as a feature.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"The corpus-based statistical parsing community has many fast and accurate automated parsing systems, including systems produced by Collins (1997), Charniak (1997) and Ratnaparkhi (1997).","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"We tabulate this increase in Table 3.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Figure 2 shows examples of extracted NE pair instances and their contexts.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"36.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"com t 700 Mountain Avenue, 2d451, Murray Hill, NJ 07974, USA.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"Such tag sequences are often treated as \u201ccomplex tags\u201d (e.g.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Other packages walk their respective data structures once to find wnf and again to find {b(wn\u22121 i )}f\u22121 i=1if necessary.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"If is a number string (such as one, two, etc), then the feature NumberString is set to 1.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"Formally, for a lexicon L and segments I \u00e2\u0088\u0088 L, O \u00e2\u0088\u0088\/ L, each word automaton accepts the language I\u00e2\u0088\u0097(O + I)I\u00e2\u0088\u0097.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"(2) was extended to have an additional, innermost loop over the (3) possible labels.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Out-of-domain test data is from the Project Syndicate web site, a compendium of political commentary.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"The relevant variables are the set of token-level tags that appear before and after each instance of the ith word type; we denote these context pairs with the set {(tb, ta)} and they are contained in t(\u00e2\u0088\u0092i).","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"The state-to-tag mapping is obtained from the best hyperparameter setting for 11 mapping shown in Table 3.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"This actually happens quite frequently (more below), so that the rankings are broad estimates.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"However, they list two sets, one consisting of 28 fragments and the other of 22 fragments, in which they had 0% recall and precision.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"First, it directly encodes linguistic intuitions about POS tag assignments: the model structure reflects the one-tag-per-word property, and a type- level tag prior captures the skew on tag assignments (e.g., there are fewer unique determiners than unique nouns).","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"Finally we show the combining techniques degrade very little when a poor parser is added to the set.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Unfortunately, we have much less data to work with than with the automatic scores.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"Although these are technically nominal, they have become known as \u00e2\u0080\u009cequational\u00e2\u0080\u009d sentences.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"So, who won the competition?","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"ATB CTB6 Negra WSJ Trees 23449 28278 20602 43948 Word Typess 40972 45245 51272 46348 Tokens 738654 782541 355096 1046829 Tags 32 34 499 45 Phrasal Cats 22 26 325 27 Test OOV 16.8% 22.2% 30.5% 13.2% Per Sentence Table 4: Gross statistics for several different treebanks.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"D o m ai n Li n k ac cu ra cy W N c o v e r a g e C C 7 3 . 3 % 2 \/ 1 1 P C 8 8 . 9 % 2 \/ 8 Table 2.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"For each extension a new position is added to the coverage set.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"nan2gual 'pumpkin.'","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"We also report word type level accuracy, the fraction of word types assigned their majority tag (where the mapping between model state and tag is determined by greedy one-to-one mapping discussed above).5 For each language, we aggregate results in the following way: First, for each hyperparameter setting, evaluate three variants: The first model (1TW) only 4 Typically, the performance stabilizes after only 10 itera-.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"In our experiments, we used the same set of features as BergKirkpatrick et al. (2010): an indicator feature based In a traditional Markov model, the emission distribution P\u0398(Xi = xi  |Zi = zi) is a set of multinomials.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"However, dynamic programming can be used to find the shortest tour in exponential time, namely in O(n22n), using the algorithm by Held and Karp.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"The text type are editorials instead of speech transcripts.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"The complexity of the quasimonotone search is O(E3 J (R2+LR)).","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"Now we have sets of phrases which share a keyword and we have links between those sets.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"Only 2 link in the CC- domain (buy-purchase, acquire-acquisition) and 2 links (trader-dealer and head-chief) in the PC- domain are found in the same synset of Word- Net 2.1 (http:\/\/wordnet.princeton.edu\/).","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"In this paper, we offer broad insight into the underperformance of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"(In this figure eps is c) be implemented, though, such as a maximal-grouping strategy (as suggested by one reviewer of this paper); or a pairwise-grouping strategy, whereby long sequences of unattached hanzi are grouped into two-hanzi words (which may have some prosodic motivation).","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"Judges where excluded from assessing the quality of MT systems that were submitted by their institution.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"00 76.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"This result suggests that all of contextual role KSs can provide useful information for resolving anaphora.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"The most popular approach to dealing with seg\u00c2\u00ad mentation ambiguities is the maximum matching method, possibly augmented with further heuristics.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"We will evaluate various specific aspects of the segmentation, as well as the overall segmentation per\u00c2\u00ad formance.","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"Instead, the designs of the various annotation layers and the actual annotation work are results of a series of diploma theses, of students\u00e2\u0080\u0099 work in course projects, and to some extent of paid assistentships.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"VBD she added VP PUNC \u00e2\u0080\u009c SBAR IN NP 0 NN.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"We start with noun features since written Arabic contains a very high proportion of NPs.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"We refer to (T , W ) as the lexicon of a language and \u00cf\u0088 for the parameters for their generation; \u00cf\u0088 depends on a single hyperparameter \u00ce\u00b2.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Within the RST \u00e2\u0080\u009cuser community\u00e2\u0080\u009d there has also been discussion whether two levels of discourse structure should not be systematically distinguished (intentional versus informational).","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"For example, a good classifier would identify Mrs. Frank as a person, Steptoe & Johnson as a company, and Honduras as a location.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"As the two NE categories are the same, we can\u00e2\u0080\u0099t differentiate phrases with different orders of par ticipants \u00e2\u0080\u0093 whether the buying company or the to-be-bought company comes first.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"This feature imposes constraints Table 1: Features based on the token string that are based on the probability of each name class during training.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"4.3 Morphological Analysis.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"The following features were used: full-string=x The full string (e.g., for Maury Cooper, full- s tring=Maury_Cooper). contains(x) If the spelling contains more than one word, this feature applies for any words that the string contains (e.g., Maury Cooper contributes two such features, contains (Maury) and contains (Cooper) . allcapl This feature appears if the spelling is a single word which is all capitals (e.g., IBM would contribute this feature). allcap2 This feature appears if the spelling is a single word which is all capitals or full periods, and contains at least one period.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"Covering the first uncovered position in the source sentence, we use the language model probability p(ej$; $).","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"The computing time, the number of search errors, and the multi-reference WER (mWER) are shown as a function of t0.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"We computed BLEU scores for each submission with a single reference translation.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Finally, other packages implement language model estimation while we are currently dependent on them to generate an ARPA file.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Besides the lack of a clear definition of what constitutes a correct segmentation for a given Chinese sentence, there is the more general issue that the test corpora used in these evaluations differ from system to system, so meaningful comparison between systems is rendered even more difficult.","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"A different notion of information structure, is used in work such as that of (?), who tried to characterize felicitous constituent ordering (theme choice, in particular) that leads to texts presenting information in a natural, \u00e2\u0080\u009cflowing\u00e2\u0080\u009d way rather than with abrupt shifts of attention.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"On each language we investigate the contribution of each component of our model.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"For example Chen and Liu (1992) report precision and recall rates of over 99%, but this counts only the words that occur in the test corpus that also occur in their dictionary.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"The horizontal axis in this plot represents the most significant dimension, which explains 62% of the variation.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"However, such an approach requires a number of hand-coded systems, which may not be available in languages other than English.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"In addition to the Europarl test set, we also collected 29 editorials from the Project Syndicate website2, which are published in all the four languages of the shared task.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"In fact, we found that enabling IRSTLM\u2019s cache made it slightly slower, so results in Table 1 use IRSTLM without caching.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"4.1 Corpora.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"The sentence length probability p(JjI) is omitted without any loss in performance.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"It rewards matches of n-gram sequences, but measures only at most indirectly overall grammatical coherence.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"In total 13,976 phrases are assigned to sets of phrases, and the accuracy on our evaluation data ranges from 65 to 99%, depending on the domain and the size of the sets.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"Additionally, for a given coverage set, at most 250 different hypotheses are kept during the search process, and the number of different words to be hypothesized by a source word is limited.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Lexical-knowledge-based approaches that include statistical information generally presume that one starts with all possible segmentations of a sentence, and picks the best segmentation from the set of possible segmentations using a probabilistic or cost\u00c2\u00ad based scoring mechanism.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"html 4 www.wagsoft.com\/RSTTool assigning rhetorical relations is a process loaded with ambiguity and, possibly, subjectivity.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"32 81.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"Thus, for example, one successor process will be have M to be in the existential state qa with the indices encoding xi , , xn, in the first 2n i tapes.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"For the English RST-annotated corpus that is made available via LDC, his corresponding result is 62%.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"Performance typically stabilizes across languages after only a few number of iterations.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"We create equivalence classes for verb, noun, and adjective POS categories.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":".., Wn ) (obs) P T : Tag assigns (T1 ,.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"Tag set As is standard, for all experiments, we set the number of latent model tag states to the size of the annotated tag set.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"The difference is due to more careful S-NOM NP NP NP VP VBG :: b NP restoring NP ADJP NN :: b NP NN NP NP ADJP DTJJ ADJP DTJJ NN :: b NP NP NP ADJP ADJP DTJJ J ..i NN :: b NP NP NP ADJP ADJP DTJJ NN _;\u00ef\u00bf\u00bd NP PRP DTJJ DTJJ J ..i _;\u00ef\u00bf\u00bd PRP J ..i NN _;\u00ef\u00bf\u00bd NP PRP DTJJ NN _;\u00ef\u00bf\u00bd NP PRP DTJJ J ..i role its constructive effective (b) Stanford (c) Berkeley (d) Bik el (a) Reference Figure 4: The constituent Restoring of its constructive and effective role parsed by the three different models (gold segmentation).","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"4 65.9 48.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"As the two NE categories are the same, we can\u00e2\u0080\u0099t differentiate phrases with different orders of par ticipants \u00e2\u0080\u0093 whether the buying company or the to-be-bought company comes first.","label":0},{"source_text":"There is no global pruning.","target_text":"Kollege.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"We report the F1 value of both measures.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"Thus corresponding pseudo-labels for instances on which gj abstain are set to zero and these instances do not contribute to the objective function.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"If 0(A) gives the number of occurrences of each terminal in the structure named by A, then, given the constraints imposed on the formalism, for each rule A --. fp(Ai, , An) we have the equality where c\u201e is some constant.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"In the pinyin transliterations a dash(-) separates syllables that may be considered part of the same phonological word; spaces are used to separate plausible phonological words; and a plus sign (+) is used, where relevant, to indicate morpheme boundaries of interest.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"C is the union of the sets of constituents suggested by the parsers. r(c) is a binary function returning t (for true) precisely when the constituent c E C should be included in the hypothesis.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"Lexicon Feature of Previous and Next Token: The string of the previous token and the next token is used with the initCaps information of . If has initCaps, then a feature (initCaps, ) is set to 1.","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"Realizing gains in practice can be challenging, however, particularly when the target domain is distant from the background data.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"Specifically, we assume each word type W consists of feature-value pairs (f, v).","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"In the natural disasters domain, agents are often forces of nature, such as hurricanes or wildfires.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"Simply assigning to each word its most frequent associated tag in a corpus achieves 94.6% accuracy on the WSJ portion of the Penn Treebank.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"Segmentation errors cascade into the parsing phase, placing an artificial limit on parsing performance.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"Proper names are assumed to be coreferent if they match exactly, or if they closely match based on a few heuristics.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"The cities of the traveling salesman problem correspond to source Table 1: DP algorithm for statistical machine translation.","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"There exist a few robust broad-coverage parsers that produce non-projective dependency structures, notably Tapanainen and J\u00a8arvinen (1997) and Wang and Harper (2004) for English, Foth et al. (2004) for German, and Holan (2004) for Czech.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"The PROBING model was designed to improve upon SRILM by using linear probing hash tables (though not arranged in a trie), allocating memory all at once (eliminating the need for full pointers), and being easy to compile.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"6 Conclusions.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"(2003), which gives 96.8% accuracy on the test set.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"Instead of offsetting new topics with punctuation, writers of MSA in sert connectives such as \u00ef\u00bf\u00bd wa and \u00ef\u00bf\u00bd fa to link new elements to both preceding clauses and the text as a whole.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"We observe similar trends when using another measure \u00e2\u0080\u0093 type-level accuracy (defined as the fraction of words correctly assigned their majority tag), according to which La ng ua ge M etr ic B K 10 E M B K 10 L B F G S G 10 F EA T S B es t F EA T S M ed ia n E ng lis h 1 1 m 1 4 8 . 3 6 8 . 1 5 6 . 0 7 5 . 5 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 0 . 9 6 6 . 4 4 7 . 8 6 6 . 4 D an is h 1 1 m 1 4 2 . 3 6 6 . 7 4 2 . 6 5 8 . 0 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 2 . 1 6 1 . 2 4 3 . 2 6 0 . 7 D ut ch 1 1 m 1 5 3 . 7 6 7 . 0 5 5 . 1 6 4 . 7 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 6 . 4 6 9 . 0 5 1 . 5 6 7 . 3 Po rtu gu es e 1 1 m 1 5 0 . 8 7 5 . 3 4 3 . 2 7 4 . 8 44 .5 69 .2 6 4 . 1 7 4 . 5 5 6 . 5 7 0 . 1 S pa ni sh 1 1 m 1 \u00e2\u0080\u0093 \u00e2\u0080\u0093 4 0 . 6 7 3 . 2 \u00e2\u0080\u0093 \u00e2\u0080\u0093 5 8 . 3 6 8 . 9 5 0 . 0 5 7 . 2 Table 4: Comparison of our method (FEATS) to state-of-the-art methods.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"Note that while the standard HMM, has O(K n) emission parameters, our model has O(n) effective parameters.3 Token Component Once HMM parameters (\u00cf\u0086, \u00ce\u00b8) have been drawn, the HMM generates a token-level corpus w in the standard way: P (w, t|\u00cf\u0086, \u00ce\u00b8) = P (T , W , \u00ce\u00b8, \u00cf\u0088, \u00cf\u0086, t, w|\u00ce\u00b1, \u00ce\u00b2) = P (T , W , \u00cf\u0088|\u00ce\u00b2) [Lexicon] \u00ef\u00a3\u00ab n n \u00ef\u00a3\u00ad (w,t)\u00e2\u0088\u0088(w,t) j \u00ef\u00a3\u00b6 P (tj |\u00cf\u0086tj\u00e2\u0088\u00921 )P (wj |tj , \u00ce\u00b8tj )\u00ef\u00a3\u00b8 P (\u00cf\u0086, \u00ce\u00b8|T , \u00ce\u00b1, \u00ce\u00b2) [Parameter] P (w, t|\u00cf\u0086, \u00ce\u00b8) [Token] We refer to the components on the right hand side as the lexicon, parameter, and token component respectively.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"The features are used to represent each example for the learning algorithm.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"It is generally considered to be fast (Pauls 29 \u2212 1 probabilities and 2' \u2212 2 non-zero backoffs. and Klein, 2011), with a default implementation based on hash tables within each trie node.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"Here, an NE instance pair is any pair of NEs separated by at most 4 syntactic chunks; for example, \u00e2\u0080\u009cIBM plans to acquire Lotus\u00e2\u0080\u009d.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"The computing time, the number of search errors, and the multi-reference WER (mWER) are shown as a function of t0.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"We excluded these from the evaluation as they can be easily identified with a list of days\/months.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"This is demonstrated by average scores over all systems, in terms of BLEU, fluency and adequacy, as displayed in Figure 5.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"In this way, the method reported on here will necessarily be similar to a greedy method, though of course not identical.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"We are not claiming that this method is almighty.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"The final block in table 2 shows models trained on feature subsets and on the SVM feature described in 3.4.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"A named entity recognizer (NER) is useful in many NLP applications such as information extraction, question answering, etc. On its own, a NER can also provide users who are looking for person or organization names with quick information.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"The focus of this work is on building POS taggers for foreign languages, assuming that we have an English POS tagger and some parallel text between the two languages.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"The number of NE instance pairs used in their experiment is less than half of our method.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"Corporate Suffixes and Person Prefixes of Other Occurrences (CSPP): If McCann has been seen as Mr. McCann somewhere else in the document, then one would like to give person a higher probability than organization.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"In such cases we assign all of the estimated probability mass to the form with the most likely pronunciation (determined by inspection), and assign a very small probability (a very high cost, arbitrarily chosen to be 40) to all other variants.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"The state in future has not enough work for its many teachers.","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"For each token, three types of features may be taken into account: the word form; the part-of-speech assigned by an automatic tagger; and labels on previously assigned dependency arcs involving the token \u2013 the arc from its head and the arcs to its leftmost and rightmost dependent, respectively.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"Note that while the standard HMM, has O(K n) emission parameters, our model has O(n) effective parameters.3 Token Component Once HMM parameters (\u00cf\u0086, \u00ce\u00b8) have been drawn, the HMM generates a token-level corpus w in the standard way: P (w, t|\u00cf\u0086, \u00ce\u00b8) = P (T , W , \u00ce\u00b8, \u00cf\u0088, \u00cf\u0086, t, w|\u00ce\u00b1, \u00ce\u00b2) = P (T , W , \u00cf\u0088|\u00ce\u00b2) [Lexicon] \u00ef\u00a3\u00ab n n \u00ef\u00a3\u00ad (w,t)\u00e2\u0088\u0088(w,t) j \u00ef\u00a3\u00b6 P (tj |\u00cf\u0086tj\u00e2\u0088\u00921 )P (wj |tj , \u00ce\u00b8tj )\u00ef\u00a3\u00b8 P (\u00cf\u0086, \u00ce\u00b8|T , \u00ce\u00b1, \u00ce\u00b2) [Parameter] P (w, t|\u00cf\u0086, \u00ce\u00b8) [Token] We refer to the components on the right hand side as the lexicon, parameter, and token component respectively.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Since we could not bias the subjects towards a particular segmentation and did not presume linguistic sophistication on their part, the instructions were simple: subjects were to mark all places they might plausibly pause if they were reading the text aloud.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"For comparison, the completely unsupervised feature-HMM baseline accuracy on the universal POS tags for English is 79.4%, and goes up to 88.7% with a treebank dictionary.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"Features and context were initially introduced into the models, but they refused to offer any gains in performance.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"The 3rd block contains the mixture baselines.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"The anaphor and antecedent appear in boldface.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"The set is then compared with the set generated from the Penn Treebank parse to determine the precision and recall.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"This is appropriate in cases where it is sanctioned by Bayes\u2019 law, such as multiplying LM and TM probabilities, but for adaptation a more suitable framework is often a mixture model in which each event may be generated from some domain.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"For locations, tokens are matched against unigrams, and sequences of two consecutive tokens are matched against bigrams.","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"As long as the main evaluation metric is dependency accuracy per word, with state-of-the-art accuracy mostly below 90%, the penalty for not handling non-projective constructions is almost negligible.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Zone: MUC data contains SGML tags, and a document is divided into zones (e.g., headlines and text zones).","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"Besides size of training data, the use of dictionaries is another factor that might affect performance.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"In this work, we take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"It falls short of the \u201cProjection\u201d baseline for German, but is statistically indistinguishable in terms of accuracy.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Coverage indicates the fraction of hypotheses in which the character yield exactly matched the reference.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"This is can not be the only explanation, since the discrepancy still holds, for instance, for out-of-domain French-English, where Systran receives among the best adequacy and fluency scores, but a worse BLEU score than all but one statistical system.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"Other errors include NE tagging errors and errors due to a phrase which includes other NEs.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"All of these systems were run on data that was not seen during their development.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"For example, from the sentence \u00e2\u0080\u009cMr.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"It is closer to the smaller value of precision and recall when there is a large skew in their values.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"When keys are longer than 64 bits, we conserve space by replacing the keys with their 64-bit hashes.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"We compared the ATB5 to tree- banks for Chinese (CTB6), German (Negra), and English (WSJ) (Table 4).","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"e0; e are the last two target words, C is a coverage set for the already covered source positions and j is the last position visited.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"(1992).","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"Step 4.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"For each co-occurrence relation (noun\/caseframe for CFLex, and caseframe\/caseframe for CFNet), BABAR computes its log-likelihood value and looks it up in the \u00cf\u00872 table to obtain a confidence level.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"We refer to different readings as different analyses whereby the segments are deterministic given the sequence of PoS tags.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"Here, the pruning threshold t0 = 10:0 is used.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"Exploiting Diversity in Natural Language Processing: Combining Parsers","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"Although embedding this version of LCFRS's in the framework of ILFP developed by Rounds (1985) is straightforward, our motivation was to capture properties shared by a family of grammatical systems and generalize them defining a class of related formalisms.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"It is difficult to motivate these days why one ministry should be exempt from cutbacks \u00e2\u0080\u0094 at the expense of the others.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"The terms on the right-hand-side denote the type-level and token-level probability terms respectively.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"Manual and Automatic Evaluation of Machine Translation between European Languages","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"Manual and Automatic Evaluation of Machine Translation between European Languages","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"(4) gives In order to minimize Zt, at each iteration the final algorithm should choose the weak hypothesis (i.e., a feature xt) which has values for W+ and W_ that minimize Equ.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"This is similar to using the Linux MAP POPULATE flag that is our default loading mechanism.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"3.3 Evaluation Results.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"We then show that the combining techniques presented above give better parsing accuracy than any of the individual parsers.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Twentieth-century linguistic work on Chinese (Chao 1968; Li and Thompson 1981; Tang 1988,1989, inter alia) has revealed the incorrectness of this traditional view.","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"The Prague Dependency Treebank (PDT) consists of more than 1M words of newspaper text, annotated on three levels, the morphological, analytical and tectogrammatical levels (Haji\u02c7c, 1998).","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"0 68.4 48.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"Even with less training data, MENERGI outperforms Borthwick' s MENE + reference resolution (Borthwick, 1999).","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"This suggests that different types of anaphora may warrant different treatment: definite NP resolution may depend more on lexical semantics, while pronoun resolution may depend more on contextual semantics.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"\u00e2\u0080\u00a2 Anaphoric links: the annotator is asked to specify whether the anaphor is a repetition, partial repetition, pronoun, epithet (e.g., Andy Warhol \u00e2\u0080\u0093 the PopArt artist), or is-a (e.g., Andy Warhol was often hunted by photographers.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"The confidence intervals are computed by bootstrap resampling for BLEU, and by standard significance testing for the manual scores, as described earlier in the paper.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"This is the only important case, because otherwise the simple majority combining technique would pick the correct constituent.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"First, it directly encodes linguistic intuitions about POS tag assignments: the model structure reflects the one-tag-per-word property, and a type- level tag prior captures the skew on tag assignments (e.g., there are fewer unique determiners than unique nouns).","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"For monolingual treebank data we relied on the CoNLL-X and CoNLL-2007 shared tasks on dependency parsing (Buchholz and Marsi, 2006; Nivre et al., 2007).","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Assigning rhetorical relations thus poses questions that can often be answered only subjectively.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"In all figures, we present the per-sentence normalized judgements.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Figure 1: Translation of PCC sample commentary (STTS)2.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"The confidence intervals are computed by bootstrap resampling for BLEU, and by standard significance testing for the manual scores, as described earlier in the paper.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"For each sentence, we counted how many n-grams in the system output also occurred in the reference translation.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"Ignoring the identity of the target language words e and e0, the possible partial hypothesis extensions due to the IBM restrictions are shown in Table 2.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"The details are given in (Tillmann, 2000).","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"rhetorical analysis We are experimenting with a hybrid statistical and knowledge-based system for discourse parsing and summarization (Stede 2003), (Hanneforth et al. 2003), again targeting the genre of commentaries.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"The accuracies for link were 73% and 86% on two evaluated domains.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"Pr(eI 1) is the language model of the target language, whereas Pr(fJ 1 jeI1) is the transla tion model.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Moreover, the Stanford parser achieves the most exact Leaf Ancestor matches and tagging accuracy that is only 0.1% below the Bikel model, which uses pre-tagged input.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"In addition to using the simple features directly, we also trained an SVM classifier with these features to distinguish between IN and OUT phrase pairs.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"This work was supported in part by the National Science Foundation under grant IRI9704240.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"A formalism generating tree sets with complex path sets can still generate only semilinear languages if its tree sets have independent paths, and semilinear path sets.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"The first stage identifies a keyword in each phrase and joins phrases with the same keyword into sets.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"All the NE pair instances which co-occur separated by at most 4 chunks are collected along with information about their NE types and the phrase between the NEs (the \u00e2\u0080\u0098context\u00e2\u0080\u0099).","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"Taking the intersection of languages in these resources, and selecting languages with large amounts of parallel data, yields the following set of eight Indo-European languages: Danish, Dutch, German, Greek, Italian, Portuguese, Spanish and Swedish.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"The probabilistic version of this procedure is straightforward: We once again assume independence among our various member parsers.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"Corpus Step 1 NE pair instances Step 2 Step 1.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Segmentation errors cascade into the parsing phase, placing an artificial limit on parsing performance.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"This significantly underperforms log-linear combination.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"We confirm the finding by Callison-Burch et al. (2006) that the rule-based system of Systran is not adequately appreciated by BLEU.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"If one is interested in translation, one would probably want to consider show up as a single dictionary word since its semantic interpretation is not trivially derivable from the meanings of show and up.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"The sign test checks, how likely a sample of better and worse BLEU scores would have been generated by two systems of equal performance.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"Let s = a + b.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"However, in practice, unknown word models also make the distribution improper.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"However, some caveats are in order in comparing this method (or any method) with other approaches to seg\u00c2\u00ad mentation reported in the literature.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","label":1},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"6. 3) all G o l d P O S 7 0 0.7 91 0.825 358 0.7 73 0.818 358 0.8 02 0.836 452 80.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"(2009) also report results on English, but on the reduced 17 tag set, which is not comparable to ours).","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Furthermore, by inverting the transducer so that it maps from phonemic transcriptions to hanzi sequences, one can apply the segmenter to other problems, such as speech recognition (Pereira, Riley, and Sproat 1994).","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"At the same time, the n-gram error rate is sensitive to samples with extreme n-gram counts.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"selected; and that recall is defined to be the number of correct hits divided by the number of items that should have been selected.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"The PROBING model is 2.4 times as fast as the fastest alternative, SRILM, and uses less memory too.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"To optimize left-to-right queries, we extend state to store backoff information: where m is the minimal context from Section 4.1 and b is the backoff penalty.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"1).","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"Manual annotation results in human in- terpretable grammars that can inform future tree- bank annotation decisions.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"The out-of-domain test set differs from the Europarl data in various ways.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"Hence, we use the bootstrap resampling method described by Koehn (2004).","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"The logistic function, whose outputs are in [0, 1], forces pp(s, t) <_ po(s, t).","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"The DL-CoTrain algorithm can be motivated as being a greedy method of satisfying the above 2 constraints.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Timing is based on plentiful memory.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Richer tag sets have been suggested for modeling morphologically complex distinctions (Diab, 2007), but we find that linguistically rich tag sets do not help parsing.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Although the segmentation requirements for Arabic are not as extreme as those for Chinese, Arabic is written with certain cliticized prepositions, pronouns, and connectives connected to adjacent words.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"The problem is a binary classification problem.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"We first make use of our morphological analyzer to find all segmentation possibilities by chopping off all prefix sequence possibilities (including the empty prefix) and construct a lattice off of them.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"Although the ultimate goal of the Verbmobil project is the translation of spoken language, the input used for the translation experiments reported on in this paper is the (more or less) correct orthographic transcription of the spoken sentences.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"\u2022 We evaluated translation from English, in addition to into English.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"3.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"2.3 Rhetorical structure.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"An important reason for separating the two types of features is that this opens up the possibility of theoretical analysis of the use of unlabeled examples.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"The F- measure score increased for both domains, reflecting a substantial increase in recall with a small decrease in precision.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"May 1995).","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"gaolbu4-gaolxing4 (hap-not-happy) 'happy?'","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Given counts cn1 where e.g. c1 is the vocabulary size, total memory consumption, in bits, is Our PROBING data structure places all n-grams of the same order into a single giant hash table.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"The next two rows are results of oracle experiments.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"In this paper, our goal has been to use the notion of LCFRS's to classify grammatical systems on the basis of their strong generative capacity.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"For instance, in the recent IWSLT evaluation, first fluency annotations were solicited (while withholding the source sentence), and then adequacy annotations.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"The label for a test example with features x is then defined as In this paper we define h(x, y) as the following function of counts seen in training data: Count(x,y) is the number of times feature x is seen with label y in training data, Count(x) = EyEy Count(x, y). a is a smoothing parameter, and k is the number of possible labels.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"A direct-mapped cache makes BerkeleyLM faster on repeated queries, but their fastest (scrolling) cached version is still slower than uncached PROBING, even on cache-friendly queries.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"BABAR applies a DempsterShafer probabilistic model to make resolutions based on evidence from the contextual role knowledge sources as well as general knowledge sources.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"logical rules, and personal names; the transitive closure of the resulting machine is then computed.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"(2) Description Source Location Names http:\/\/www.timeanddate.com http:\/\/www.cityguide.travel-guides.com http:\/\/www.worldtravelguide.net Corporate Names http:\/\/www.fmlx.com Person First Names http:\/\/www.census.gov\/genealogy\/names Person Last Names Table 2: Sources of Dictionaries The McCann family . . .","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Their default variant implements a forward trie, in which words are looked up in their natural left-to-right order.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"Overview of the method 2.2 Step by Step Algorithm.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Various segmentation approaches were then compared with human performance: 1.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"The terrorism examples reflect fairly obvious relationships: people who are murdered are killed; agents that \u00e2\u0080\u009creport\u00e2\u0080\u009d things also \u00e2\u0080\u009cadd\u00e2\u0080\u009d and \u00e2\u0080\u009cstate\u00e2\u0080\u009d things; crimes that are \u00e2\u0080\u009cperpetrated\u00e2\u0080\u009d are often later \u00e2\u0080\u009ccondemned\u00e2\u0080\u009d.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Making the ten judgements (2 types for 5 systems) takes on average 2 minutes.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"4 69.0 51.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Unexpectedly, because the ministries of treasury and education both had prepared the teacher plan together.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"We are currently working on a complete open source implementation of a training and decoding system, which should become available over the summer. pus, from which also the in-domain test set is taken.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"However, in practice, unknown word models also make the distribution improper.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"The rationale for treating these semantic labels differently is that they are specific and reliable (as opposed to the WordNet classes, which are more coarse and more noisy due to polysemy).","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"For instance, in the recent IWSLT evaluation, first fluency annotations were solicited (while withholding the source sentence), and then adequacy annotations.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"While building a machine translation system is a serious undertaking, in future we hope to attract more newcomers to the field by keeping the barrier of entry as low as possible.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"We consider the unsupervised POS induction problem without the use of a tagging dictionary.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"The token-level term is similar to the standard HMM sampling equations found in Johnson (2007).","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"We asked participants to each judge 200\u2013300 sentences in terms of fluency and adequacy, the most commonly used manual evaluation metrics.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Note that Chang, Chen, and Chen (1991), in addition to word-frequency information, include a constraint-satisfication model, so their method is really a hybrid approach.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"All our results are obtained by using only the official training data provided by the MUC conferences.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"2.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"0 X u} \"' o; .2 X X><X X XX X X X X X X x X X X X X x X V X X X X .;t'*- XXX:OX X X X X X X 9 x X X XX XX X X X X X X X XXX:< X X>O<XX>!KXX XI<>< \u00c2\u00bbC X X XX :X: X X \"' X X XX >OO<X>D<XIK X X X X X X --XX\u00c2\u00bb: XXX X X\u00c2\u00bbC X X\u00c2\u00abX...C:XXX X Xll< X X ><XX>IIC:liiC:oiiiiCI--8!X:liiOC!I!S8K X X X 10 100 1000 10000 log(F)_base: R\"2=0.20 (p < 0.005) X 100000 Figure 6 Plot of log frequency of base noun, against log frequency of plural nouns.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"We compare three hash tables: our probing implementation, GCC\u2019s hash set, and Boost\u2019s8 unordered.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"For example, co-occurring caseframes may reflect synonymy (e.g., \u00e2\u0080\u009c<patient> kidnapped\u00e2\u0080\u009d and \u00e2\u0080\u009c<patient> abducted\u00e2\u0080\u009d) or related events (e.g., \u00e2\u0080\u009c<patient> kidnapped\u00e2\u0080\u009d and \u00e2\u0080\u009c<patient> released\u00e2\u0080\u009d).","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"(2006) developed a technique for splitting and chunking long sentences.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"The second row represents the performance of the median hyperparameter setting.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"For one thing, it is not clear who is to receive settlements or what should happen in case not enough teachers accept the offer of early retirement.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"We do not experiment with models larger than physical memory in this paper because TPT is unreleased, factors such as disk speed are hard to replicate, and in such situations we recommend switching to a more compact representation, such as RandLM.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"The judgements tend to be done more in form of a ranking of the different systems.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"A simple extension will be used to handle this problem.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"Using a treebank grammar, a data-driven lexicon, and a linguistically motivated unknown-tokens handling technique our model outperforms previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"To summarize, we provided: The performance of the baseline system is similar to the best submissions in last year\u2019s shared task.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"Out-of-domain test data is from the Project Syndicate web site, a compendium of political commentary.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"na me =>2 ha nzi fa mi ly 1 ha nzi gi ve n 6.1 ha nzi fa mi ly => ha nz ii 7.2 ha nzi fa mi ly => ha nzi i ha nz ij 8.1 ha nzi gi ve n => ha nz ii 9.2 ha nzi giv en => ha nzi i ha nz ij The difficulty is that given names can consist, in principle, of any hanzi or pair of hanzi, so the possible given names are limited only by the total number of hanzi, though some hanzi are certainly far more likely than others.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"Unfortunately, we have much less data to work with than with the automatic scores.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"The machine learning community has been in a similar situation and has studied the combination of multiple classifiers (Wolpert, 1992; Heath et al., 1996).","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"We call this approach parse hybridization.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"Beyond optimizing the memory size of TRIE, there are alternative data structures such as those in Guthrie and Hepple (2010).","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Since our destructor is an efficient call to munmap, bypassing the destructor favors only other packages.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"Such a classification can be seen as a not-always-correct summary of global features.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"Systems that generally do worse than others will receive a negative one.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"The second row is the accuracy of the best of the three parsers.'","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"The differences in difficulty are better reflected in the BLEU scores than in the raw un-normalized manual judgements.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"This step ensures that the most frequent terms for each domain are labeled (in case some of them are not in WordNet) and labeled with the sense most appropriate for the domain.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Presenting the output of several system allows the human judge to make more informed judgements, contrasting the quality of the different systems.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Note that it is in precision that our over\u00c2\u00ad all performance would appear to be poorer than the reported performance of Chang et al., yet based on their published examples, our system appears to be doing better precisionwise.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"As shown in Table 3, the proportion of sentences containing some non-projective dependency ranges from about 15% in DDT to almost 25% in PDT.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"The corpus was wordaligned using both HMM and IBM2 models, and the phrase table was the union of phrases extracted from these separate alignments, with a length limit of 7.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"We then gather all phrases with the same keyword.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"Removing the leaves from the resulting tree yields a parse for L under G, with the desired probabilities.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"If is not initCaps, then (not-initCaps, ) is set to 1.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"Text generation, or at least the two phases of text planning and sentence planning, is a process driven partly by well-motivated choices (e.g., use this lexeme X rather than that more colloquial near-synonym Y ) and partly by con tation like that of PCC can be exploited to look for correlations in particular between syntactic structure, choice of referring expressions, and sentence-internal information structure.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"Our full model outperforms the \u201cNo LP\u201d setting because it has better vocabulary coverage and allows the extraction of a larger set of constraint features.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"While RST (Mann, Thompson 1988) proposed that a single relation hold between adjacent text segments, SDRT (Asher, Lascarides 2003) maintains that multiple relations may hold simultaneously.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"The NE tagger is a rule-based system with 140 NE categories [Sekine et al. 2004].","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"Segmental morphology Hebrew consists of seven particles m(\u201cfrom\u201d) f(\u201cwhen\u201d\/\u201cwho\u201d\/\u201cthat\u201d) h(\u201cthe\u201d) w(\u201cand\u201d) k(\u201clike\u201d) l(\u201cto\u201d) and b(\u201cin\u201d). which may never appear in isolation and must always attach as prefixes to the following open-class category item we refer to as stem.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"na me =>1 ha nzi fa mi ly 1 ha nzi gi ve n 4.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Clearly the percentage of productively formed words is quite small (for this particular corpus), meaning that dictionary entries are covering most of the 15 GR is .73 or 96%..","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"The sentence-selection approach is crude in that it imposes a binary distinction between useful and non-useful parts of OUT.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"It is well know that language pairs such as EnglishGerman pose more challenges to machine translation systems than language pairs such as FrenchEnglish.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"In this particular case, all English vertices are labeled as nouns by the supervised tagger.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"The implementation is In a similar effort, (G\u00c2\u00a8otze 2003) developed a proposal for the theory-neutral annotation of information structure (IS) \u00e2\u0080\u0094 a notoriously difficult area with plenty of conflicting and overlapping terminological conceptions.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Further, we report current resident memory and peak virtual memory because these are the most applicable statistics provided by the kernel.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Memory usage is likely much lower than ours. fThe original paper (Germann et al., 2009) provided only 2s of query timing and compared with SRI when it exceeded available RAM.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"This is the form of recursive levels in iDafa constructs.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Microsoft\u2019s approach uses dependency trees, others use hierarchical phrase models.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"(b) After they were released...","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"Annotation consistency is important in any supervised learning task.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"Then, moving from connective to connective, ConAno sometimes offers suggestions for its scope (using heuristics like \u00e2\u0080\u0098for sub- junctor, mark all words up to the next comma as the first segment\u00e2\u0080\u0099), which the annotator can accept with a mouseclick or overwrite, marking instead the correct scope with the mouse.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"This section measures performance on shared tasks in order of increasing complexity: sparse lookups, evaluating perplexity of a large file, and translation with Moses.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"These conaUses lossy compression. bThe 8-bit quantized variant returned incorrect probabilities as explained in Section 3.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"In most cases, however, these expansions come with a steep increase in model complexity, with respect to training procedure and inference time.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"With the exception of the Dutch data set, no other processing is performed on the annotated tags.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"Such ambiguities cause discrepancies between token boundaries (indexed as white spaces) and constituent boundaries (imposed by syntactic categories) with respect to a surface form.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"This is because different judges focused on different language pairs.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"J:j:l :zhongl :0.0 ;m,Jlong4 :0.0 (mHHaryg9tltHBI) \u00c2\u00a3: _ADV: 5.98 \u00c2\u00a5 :hua2:o.o E :_NC: 4.41 :mln2:o.o mm : guo2 : 0.0 (RopubllcofChlna) .....,.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"For the \u00e2\u0080\u0098core\u00e2\u0080\u0099 portion of PCC, we found that on average, 35% of the coherence relations in our RST annotations are explicitly signalled by a lexical connective.6 When adding the fact that connectives are often ambiguous, one has to conclude that prospects for an automatic analysis of rhetorical structure using shallow methods (i.e., relying largely on connectives) are not bright \u00e2\u0080\u0094 but see Sections 3.2 and 3.3 below.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"It did 402 queries\/ms using 1.80 GB. cMemory use increased during scoring due to batch processing (MIT) or caching (Rand).","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"On a set of 11 sentence fragments-the A set-where they reported 100% recall and precision for name identification, we had 73% recall and 80% precision.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"The natural baseline (baseline) outperforms the pure IN system only for EMEA\/EP fren.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"(2006) developed a technique for splitting and chunking long sentences.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"(2009).","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"The commentaries in PCC are all of roughly the same length, ranging from 8 to 10 sentences.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"Je voudrais pr\u00b4eciser, a` l\u2019adresse du commissaire Liikanen, qu\u2019il n\u2019est pas ais\u00b4e de recourir aux tribunaux nationaux.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"Given around 90,000 unlabeled examples, the methods described in this paper classify names with over 91% accuracy.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"Asking the annotator to also formulate the question is a way of arriving at more reproducible decisions.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"Even with less training data, MENERGI outperforms Borthwick' s MENE + reference resolution (Borthwick, 1999).","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Unsupervised Learning of Contextual Role Knowledge for Coreference Resolution","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"To check for this, we do pairwise bootstrap resampling: Again, we repeatedly sample sets of sentences, this time from both systems, and compare their BLEU scores on these sets.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"The set of n-grams appearing in a model is sparse, and we want to efficiently find their associated probabilities and backoff penalties.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"1 1 0.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"The correct resolution in sentence (b) comes from knowledge that people who are kidnapped are often subsequently released.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"These results are promising and there are several avenues for improving on these results.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"The developer explained that the loading process requires extra memory that it then frees. eBased on the ratio to SRI\u2019s speed reported in Guthrie and Hepple (2010) under different conditions.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"The choice of the particular newspaper was motivated by the fact that the language used in a regional daily is somewhat simpler than that of papers read nationwide.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"During testing, it is possible that the classifier produces a sequence of inadmissible classes (e.g., person begin followed by location unique).","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"We would like to thank Eugene Charniak, Michael Collins, and Adwait Ratnaparkhi for enabling all of this research by providing us with their parsers and helpful comments.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"(Hartmann 1984), for example, used the term Reliefgebung to characterize the distibution of main and minor information in texts (similar to the notion of nuclearity in RST).","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"We refer to (T , W ) as the lexicon of a language and \u00cf\u0088 for the parameters for their generation; \u00cf\u0088 depends on a single hyperparameter \u00ce\u00b2.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"including Third Tone Sandhi (Shih 1986), which changes a 3 (low) tone into a 2 (rising) tone before another 3 tone: 'j\";gil, xiao3 [lao3 shu3] 'little rat,' becomes xiao3 { lao2shu3 ], rather than xiao2 { lao2shu3 ], because the rule first applies within the word lao3shu3 'rat,' blocking its phrasal application.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"Figure 4 shows some such phrase sets based on keywords in the CC-domain.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"They considered a semi-supervised POS tagging scenario and showed that one can use a graph over trigram types, and edge weights based on distributional similarity, to improve a supervised conditional random field tagger.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"Set the decision list to include all rules whose (smoothed) strength is above some threshold Pmin.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"A morphological analyzer M : W\u2014* L is a function mapping sentences in Hebrew (W E W) to their corresponding lattices (M(W) = L E L).","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"suffixes (e.g., \u00ef\u00bf\u00bd=) Other notable parameters are second order vertical Markovization and marking of unary rules.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"We describe a novel approach for inducing unsupervised part-of-speech taggers for languages that have no labeled training data, but have translated text in a resource-rich language.","label":1},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"(Carlson, Marcu 2001) responded to this situation with relatively precise (and therefore long!)","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"While automatic measures are an invaluable tool for the day-to-day development of machine translation systems, they are only a imperfect substitute for human assessment of translation quality, or as the acronym BLEU puts it, a bilingual evaluation understudy.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"The second term is a regularizer and encourages all type marginals to be uniform to the extent that is allowed by the first two terms (cf. maximum entropy principle).","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"We show how a datadriven deterministic dependency parser, in itself restricted to projective structures, can be combined with graph transformation techniques to produce non-projective structures.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"6.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"But diacritics are not present in unvocalized text, which is the standard form of, e.g., news media documents.3 VBD she added VP PUNC S VP VBP NP ...","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"Cur In order to ground such approaches in linguistic observation and description, a multi-level anno 10 For an exposition of the idea as applied to the task of text planning, see (Chiarcos, Stede 2004).","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"We feel that information from a whole corpus might turn out to be noisy if the documents in the corpus are not of the same genre.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"Based on these high-confidence alignments we can extract tuples of the form [u H v], where u is a foreign trigram type, whose middle word aligns to an English word type v. Our bilingual similarity function then sets the edge weights in proportion to these tuple counts.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"An important reason for separating the two types of features is that this opens up the possibility of theoretical analysis of the use of unlabeled examples.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"We consider the unsupervised POS induction problem without the use of a tagging dictionary.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"com t 600 Mountain Avenue, 2c278, Murray Hill, NJ 07974, USA.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"One striking example is the error reduction for Spanish, which reduces error by 36.5% and 24.7% for the best and median settings respectively.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"We picked two domains, the CC-domain and the \u00e2\u0080\u009cPerson \u00e2\u0080\u0093 Company\u00e2\u0080\u009d domain (PC-domain), for the evaluation, as the entire system output was too large to evaluate.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"Finally, we thank Kuzman Ganchev and the three anonymous reviewers for helpful suggestions and comments on earlier drafts of this paper.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"They are also labelled for their topicality (yes \/ no), and this annotation is accompanied by a confidence value assigned by the annotator (since it is a more subjective matter).","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"The accuracies for link were 73% and 86% on two evaluated domains.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"D o m ai n # of ph ras es t o t a l p h r a s e s ac cu ra cy C C 7 o r m o r e 1 0 5 8 7 . 6 % 6 o r l e s s 1 0 6 6 7 . 0 % P C 7 o r m o r e 3 5 9 9 9 . 2 % 6 o r l e s s 2 5 5 6 5 . 1 % Table 1.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"Although matching is done at the sentence level, this information is subsequently discarded when all matches are pooled.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"Recent work by Finkel and Manning (2009) which re-casts Daum\u00b4e\u2019s approach in a hierarchical MAP framework may be applicable to this problem.","label":0},{"source_text":"There is no global pruning.","target_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"German English Training: Sentences 58 073 Words 519 523 549 921 Words* 418 979 453 632 Vocabulary Size 7939 4648 Singletons 3454 1699 Test-147: Sentences 147 Words 1 968 2 173 Perplexity { 26:5 Table 4: Multi-reference word error rate (mWER) and subjective sentence error rate (SSER) for three different search procedures.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"Although these are technically nominal, they have become known as \u00e2\u0080\u009cequational\u00e2\u0080\u009d sentences.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"36.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"To prevent this we &quot;smooth&quot; the confidence by adding a small value, e, to both W+ and W_, giving at = Plugging the value of at from Equ.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"We combined evidence from four contextual role knowledge sources with evidence from seven general knowledge sources using a DempsterShafer probabilistic model.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"The method being described-henceforth ST..","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"As illustrated by MCTAG's, it is possible for a formalism to give tree sets with bounded dependent paths while still sharing the constrained rewriting properties of CFG's, HG's, and TAG's.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"98 15.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"The state in future has not enough work for its many teachers.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"Two aspects of the corpus have been presented in previous papers ((Re- itter, Stede 2003) on underspecified rhetorical structure; (Stede 2003) on the perspective of knowledge-based summarization).","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"Lattice parsing (Chappelier et al., 1999) is an alternative to a pipeline that prevents cascading errors by placing all segmentation options into the parse chart.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"First, it directly encodes linguistic intuitions about POS tag assignments: the model structure reflects the one-tag-per-word property, and a type- level tag prior captures the skew on tag assignments (e.g., there are fewer unique determiners than unique nouns).","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"\u2022 We evaluated translation from English, in addition to into English.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"92 76.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"37 84.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Nonetheless, parse quality is much lower in the joint model because a lattice is effectively a long sentence.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"Thus, provided at least this amount of IN data is available\u2014as it is in our setting\u2014adapting these weights is straightforward.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"\u00e2\u0080\u00a2 Some tools would allow for the desired annotation mode, but are so complicated (they can be used for many other purposes as well) that annotators take a long time getting used to them.","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"As each global feature group is added to the list of features, we see improvements to both MUC6 and MUC6 MUC7 Baseline 90.75% 85.22% + ICOC 91.50% 86.24% + CSPP 92.89% 86.96% + ACRO 93.04% 86.99% + SOIC 93.25% 87.22% + UNIQ 93.27% 87.24% Table 3: F-measure after successive addition of each global feature group Table 5: Comparison of results for MUC6 Systems MUC6 MUC7 No.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"Restrictions We compare our new approach with the word reordering used in the IBM translation approach (Berger et al., 1996).","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"For illustration, an English translation of one of the commentaries is given in Figure 1.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"Ablation Analysis We evaluate the impact of incorporating various linguistic features into our model in Table 3.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"BABAR uses information extraction patterns to identify contextual roles and creates four contextual role knowledge sources using unsupervised learning.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"In other words, the set of hidden states F was chosen to be the fine set of treebank tags.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"All three curves remain steep at the maximum training set size of 18818 trees.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"This departure from the traditional token-based tagging approach allows us to explicitly capture type- level distributional properties of valid POS tag as signments as part of the model.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"For example, the Wang, Li, and Chang system fails on the sequence 1:f:p:]nian2 nei4 sa3 in (k) since 1F nian2 is a possible, but rare, family name, which also happens to be written the same as the very common word meaning 'year.'","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"Our annotators pointed out that very often they made almost random decisions as to what relation to choose, and where to locate the boundary of a span.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"As can be seen in the example, the first two phrases have a different order of NE names from the last two, so we can determine that the last two phrases represent a reversed relation.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"We loosely describe the class of all such systems as Linear Context-Free Rewriting Formalisms.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"For every sequence of initial capitalized words, its longest substring that occurs in the same document as a sequence of initCaps is identified.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"As we have said, parse quality decreases with sentence length.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"A graphical depiction of our model as well as a summary of random variables and parameters can be found in Figure 1.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"Find keywords for each NE pair When we look at the contexts for each domain, we noticed that there is one or a few important words which indicate the relation between the NEs (for example, the word \u00e2\u0080\u009cunit\u00e2\u0080\u009d for the phrase \u00e2\u0080\u009ca unit of\u00e2\u0080\u009d).","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Given the closeness of most systems and the wide over-lapping confidence intervals it is hard to make strong statements about the correlation between human judgements and automatic scoring methods such as BLEU.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"BerkeleyLM revision 152 (Pauls and Klein, 2011) implements tries based on hash tables and sorted arrays in Java with lossy quantization.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Sparse lookup is a key subproblem of language model queries.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"A possible probabilistic model for assigning probabilities to complex analyses of a surface form may be and indeed recent sequential disambiguation models for Hebrew (Adler and Elhadad, 2006) and Arabic (Smith et al., 2005) present similar models.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"Fortunately, performance was stable across various values, and we were able to use the same hyperparameters for all languages.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"Upon identifying an anaphoric expression (currently restricted to: pronouns, prepositional adverbs, definite noun phrases), the an- notator first marks the antecedent expression (currently restricted to: various kinds of noun phrases, prepositional phrases, verb phrases, sentences) and then establishes the link between the two.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"But Arabic contains a variety of linguistic phenomena unseen in English.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"We can show that languages generated by LCFRS's are semilinear as long as the composition operation does not remove any terminal symbols from its arguments.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"5 We choose these two metrics over the Variation Information measure due to the deficiencies discussed in Gao and Johnson (2008).","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"Each word is simply tagged with the semantic classes corresponding to all of its senses.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"We check, how likely only up to k = 20 better scores out of n = 100 would have been generated by two equal systems, using the binomial distribution: If p(0..k; n, p) < 0.05, or p(0..k; n, p) > 0.95 then we have a statistically significant difference between the systems.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"This approach needs a phrase as an initial seed and thus the possible relationships to be extracted are naturally limited.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"Tag set As is standard, for all experiments, we set the number of latent model tag states to the size of the annotated tag set.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"5 70.1 58.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"For example, in Figure 3, we can see that the phrases in the \u00e2\u0080\u009cbuy\u00e2\u0080\u009d, \u00e2\u0080\u009cacquire\u00e2\u0080\u009d and \u00e2\u0080\u009cpurchase\u00e2\u0080\u009d sets are mostly paraphrases.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Let us consider an example of ambiguity caused by devocalization.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"Because all English vertices are going to be labeled, we do not need to disambiguate them by embedding them in trigrams.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"For example, even if the contexts surrounding an anaphor and candidate match exactly, they are not coreferent if they have substantially different meanings 9 We would be happy to make our manually annotated test data available to others who also want to evaluate their coreference resolver on the MUC4 or Reuters collections.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Our results show that BABAR achieves good performance in both domains, and that the contextual role knowledge improves performance, especially on pronouns.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"It is likely that the more complex cases, where path information could make a difference, are beyond the reach of the parser in most cases.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"The bias of automatic methods in favor of statistical systems seems to be less pronounced on out-of-domain test data.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"Ex: The brigade, which attacked ...","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"The simplest version of the maximum matching algorithm effectively deals with ambiguity by ignoring it, since the method is guaranteed to produce only one segmentation.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"Model components cascade, so the row corresponding to +FEATS also includes the PRIOR component (see Section 3).","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"At most one feature in this group will be set to 1.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"The high 1 tone of J1l would not normally neutralize in this fashion if it were functioning as a word on its own.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"The inventions disclosed herein are the subject of a patent application owned by the University of Utah and licensed on an exclusive basis to Attensity Corporation.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"However, until such standards are universally adopted in evaluating Chinese segmenters, claims about performance in terms of simple measures like percent correct should be taken with a grain of salt; see, again, Wu and Fung (1994) for further arguments supporting this conclusion.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"This task measures how well each package performs in machine translation.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"The German finite verbs 'bin' (second example) and 'konnten' (third example) are too far away from the personal pronouns 'ich' and 'Sie' (6 respectively 5 source sentence positions).","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Hash tables are a common sparse mapping technique used by SRILM\u2019s default and BerkeleyLM\u2019s hashed variant.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"The work of Rounds (1969) shows that the path sets of trees derived by IG's (like those of TAG's) are context-free languages.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"splitIN captures the verb\/preposition idioms that are widespread in Arabic.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Part of the gap between resident and virtual memory is due to the time at which data was collected.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"On the other hand, when all systems produce muddled output, but one is better, and one is worse, but not completely wrong, a judge is inclined to hand out judgements of 4, 3, and 2.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"This drastic tree manipulation is not appropriate for situations in which we want to assign particular structures to sentences.","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"Corpus Step 1 NE pair instances Step 2 Step 1.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"We tabulate this increase in Table 3.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"A moment's reflection will reveal that things are not quite that simple.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"It reads a file with a list of German connectives, and when a text is opened for annotation, it highlights all the words that show up in this list; these will be all the potential connectives.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"As can be seen in the example, the first two phrases have a different order of NE names from the last two, so we can determine that the last two phrases represent a reversed relation.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"The corpora for both settings are summarized in table 1.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Lexical-knowledge-based approaches that include statistical information generally presume that one starts with all possible segmentations of a sentence, and picks the best segmentation from the set of possible segmentations using a probabilistic or cost\u00c2\u00ad based scoring mechanism.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"In general, m; l; l0 6= fl1; l2; l3g and in line umber 3 and 4, l0 must be chosen not to violate the above reordering restriction.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"This paper proposes a simple and effective tagging method that directly models tag sparsity and other distributional properties of valid POS tag assignments.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"We contrasted formalisms such as CFG's, HG's, TAG's and MCTAG's, with formalisms such as IG's and unificational systems such as LFG's and FUG's.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"2 70.7 52.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"For each label (Per s on, organization and Location), take the n contextual rules with the highest value of Count' (x) whose unsmoothed3 strength is above some threshold pmin.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"Hence, for each token, one of the four features zone-TXT, zone- HL, zone-DATELINE, or zone-DD is set to 1, and the other 3 are set to 0.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"This WFST is then summed with the WFST implementing the dictionary and morphological rules, and the transitive closure of the resulting transducer is computed; see Pereira, Riley, and Sproat (1994) for an explanation of the notion of summing WFSTs.12 Conceptual Improvements over Chang et al.'s Model.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"Sentence (2) and (3) help to disambiguate one way or the other.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"This is can not be the only explanation, since the discrepancy still holds, for instance, for out-of-domain French-English, where Systran receives among the best adequacy and fluency scores, but a worse BLEU score than all but one statistical system.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"The segmenter will give both analyses :1 cai2 neng2 'just be able,' and ?]cai2neng2 'talent,' but the latter analysis is preferred since splitting these two morphemes is generally more costly than grouping them.","label":0},{"source_text":"However, these approaches are ill-equipped to directly represent type-based constraints such as sparsity.","target_text":"See Section 5.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Hieu Hoang named the code \u201cKenLM\u201d and assisted with Moses along with Barry Haddow.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"and \u00e2\u0080\u009cH\u00e2\u0080\u009d represents \u00e2\u0080\u009cHanson Plc\u00e2\u0080\u009d.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"For some language pairs (such as GermanEnglish) system performance is more divergent than for others (such as English-French), at least as measured by BLEU.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"Cohen and Smith (2007) followed up on these results and proposed a system for joint inference of morphological and syntactic structures using factored models each designed and trained on its own.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"Microsoft\u2019s approach uses dependency trees, others use hierarchical phrase models.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"When this feature type was included, CoBoost chose this default feature at an early iteration, thereby giving non-abstaining pseudo-labels for all examples, with eventual convergence to the two classifiers agreeing by assigning the same label to almost all examples.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"A different notion of information structure, is used in work such as that of (?), who tried to characterize felicitous constituent ordering (theme choice, in particular) that leads to texts presenting information in a natural, \u00e2\u0080\u009cflowing\u00e2\u0080\u009d way rather than with abrupt shifts of attention.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"Finally, as\u00c2\u00ad suming a simple bigram backoff model, we can derive the probability estimate for the particular unseen word i\u00c2\u00a51J1l.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"Covering the first uncovered position in the source sentence, we use the language model probability p(ej$; $).","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"A summary of the corpus used in the experiments is given in Table 3.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"With some minor API changes, namely returning the length of the n-gram matched, it could also be faster\u2014though this would be at the expense of an optimization we explain in Section 4.1.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"For example, as Gan (1994) has noted, one can construct examples where the segmen\u00c2\u00ad tation is locally ambiguous but can be determined on the basis of sentential or even discourse context.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"In various dialects of Mandarin certain phonetic rules apply at the word.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Initially, the DempsterShafer model assumes that all hypotheses are equally likely, so it creates a set called \u00ce\u00b8 that includes all hypotheses.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Nicola Bertoldi and Marcello Federico assisted with IRSTLM.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"Each feature concept is akin to a random variable and its occurrence in the text corresponds to a particular instantiation of that random variable.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"Next, for each pair of NE categories, we collect all the contexts and find the keywords which are topical for that NE category pair.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Annotators argued for the importance of having correct and even multiple references.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Indeed, as we shall show in Section 5, even human judges differ when presented with the task of segmenting a text into words, so a definition of the criteria used to determine that a given segmentation is correct is crucial before one can interpret such measures.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"Overall, it gives improvements ranging from 1.1% for German to 14.7% for Italian, for an average improvement of 8.3% over the unsupervised feature-HMM model.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"A morphological analyzer M : W\u2014* L is a function mapping sentences in Hebrew (W E W) to their corresponding lattices (M(W) = L E L).","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"Zt can be written as follows Following the derivation of Schapire and Singer, providing that W+ > W_, Equ.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"We model po(s|t) using a MAP criterion over weighted phrase-pair counts: and from the similarity to (5), assuming y = 0, we see that w\u03bb(s, t) can be interpreted as approximating pf(s, t)\/po(s, t).","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"We tried two versions of our graph-based approach: feature after the first stage of label propagation (Eq.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"And time is short.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"A search restriction especially useful for the translation direction from German to English is presented.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Finally, we provide a realistic eval uation in which segmentation is performed both in a pipeline and jointly with parsing (\u00c2\u00a76).","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"The judgement of 4 in the first case will go to a vastly better system output than in the second case.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"Each pattern represents the role that a noun phrase plays in the surrounding context.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"1 The apparent difficulty of adapting constituency models to non-configurational languages has been one motivation for dependency representations (Hajic\u00cb\u0087 and Zema\u00c2\u00b4nek, 2004; Habash and Roth, 2009).","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"Because of their size, the examples (Figures 2 to 4) appear at the end of the paper.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"MITLM 0.4 (Hsu and Glass, 2008) is mostly designed for accurate model estimation, but can also compute perplexity.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"From Thatcher's (1973) work, it is obvious that the complexity of the set of paths from root to frontier of trees in a local set (the tree set of a CFG) is regular'.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"We applied the AutoSlog system (Riloff, 1996) to our unannotated training texts to generate a set of extraction patterns for each domain.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"From here on we will refer to the named-entity string itself as the spelling of the entity, and the contextual predicate as the context.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"Although the tag distributions of the foreign words (Eq.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"Hence, there is a total of 29 classes (7 name classes 4 sub-classes 1 not-a-name class).","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"0 57.2 43.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"Note, however, that there might be situations in which Zco in fact increases.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"The work of the second author as well as collaboration visits to Israel was financed by NWO, grant number 017.001.271.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"We have shown the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"In this case nonalpha is the string formed by removing all upper\/lower case letters from the spelling (e.g., for Thomas E. Petry nonalpha= .","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"We respond to this on the one hand with a format for its underspecification (see 2.4) and on the other hand with an additional level of annotation that attends only to connectives and their scopes (see 2.5), which is intended as an intermediate step on the long road towards a systematic and objective treatment of rhetorical structure.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Each name class is subdivided into 4 sub-classes, i.e., N begin, N continue, N end, and N unique.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"For all languages we do not make use of a tagging dictionary.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"We do not show the numbers for the Bayes models in Table 2 because the parameters involved were established using this set.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"The first author is supported by a National Defense Science and Engineering Graduate (NDSEG) fellowship.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"88,962 (spelling,context) pairs were extracted as training data.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"We have used the Java-based opennlp maximum entropy package1.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"Judges where excluded from assessing the quality of MT systems that were submitted by their institution.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"The parser switching oracle is the upper bound on the accuracy that can be achieved on this set in the parser switching framework.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"We aligned the texts at a sentence level across all four languages, resulting in 1064 sentence per language.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"It also does not prune, so comparing to our pruned model would be unfair.","label":0},{"source_text":"The TRIE model uses less memory than the smallest lossless alternative and is still faster than SRILM.","target_text":"The returned state s(wn1) may then be used in a followon query p(wn+1js(wn1)) that extends the previous query by one word.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"Using a treebank grammar, a data-driven lexicon, and a linguistically motivated unknown-tokens handling technique our model outperforms previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"The paper explains the design decisions taken in the annotations, and describes a number of applications using this corpus with its multi-layer annotation.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"The \u201cNo LP\u201d model does not outperform direct projection for German and Greek, but performs better for six out of eight languages.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Interpolation search has a more expensive pivot but performs less pivoting and reads, so it is slow on small data and faster on large data.","label":0},{"source_text":"Nevertheless, only a part of this corpus (10 texts), which the authors name \"core corpus\", is annotated with all this information.","target_text":"The portions of information in the large window can be individually clicked visible or invisible; here we have chosen to see (from top to bottom) \u00e2\u0080\u00a2 the full text, \u00e2\u0080\u00a2 the annotation values for the activated annotation set (co-reference), \u00e2\u0080\u00a2 the actual annotation tiers, and \u00e2\u0080\u00a2 the portion of text currently \u00e2\u0080\u0098in focus\u00e2\u0080\u0099 (which also appears underlined in the full text).","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"MADA uses an ensemble of SVMs to first re-rank the output of a deterministic morphological analyzer.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"wo rd => na m e 2.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"The SynRole KS computes the relative frequency with which the candidates\u00e2\u0080\u0099 syntactic role (subject, direct object, PP object) appeared in resolutions in the training set.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"Various verbal (e.g., \u00ef\u00bf\u00bd, .::) and adjectival.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"The second main result is that the pseudo-projective approach to parsing (using special arc labels to guide an inverse transformation) gives a further improvement of about one percentage point on attachment score.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"In such cases we assign all of the estimated probability mass to the form with the most likely pronunciation (determined by inspection), and assign a very small probability (a very high cost, arbitrarily chosen to be 40) to all other variants.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"Future extensions of the system might include: 1) An extended translation model, where we use more context to predict a source word.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"This technique has the advantage of requiring no training, but it has the disadvantage of treating all parsers equally even though they may have differing accuracies or may specialize in modeling different phenomena.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"30 16.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"Memory-based classifiers for the experiments were created using TiMBL (Daelemans et al., 2003).","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"It did 402 queries\/ms using 1.80 GB. cMemory use increased during scoring due to batch processing (MIT) or caching (Rand).","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"In Section 3, we introduce our novel concept to word reordering and a DP-based search, which is especially suitable for the translation direction from German to English.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"The best analysis of the corpus is taken to be the true analysis, the frequencies are re-estimated, and the algorithm is repeated until it converges.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"MENE (without the help of hand-coded systems) has been shown to be somewhat inferior in performance.","label":0},{"source_text":"There is no global pruning.","target_text":"Restrictions: Quasi-monotone Search The above search space is still too large to allow the translation of a medium length input sentence.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"(S2) The burglar fired the gun three times and fled.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"F1 85 Berkeley 80 Stanford.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"0750271 and by the DARPA GALE program.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"This decreases the statistical significance of our results compared to those studies.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"In many cases, there is an even stronger restriction: over large portions of the source string, the alignment is monotone.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"html 4 www.wagsoft.com\/RSTTool assigning rhetorical relations is a process loaded with ambiguity and, possibly, subjectivity.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"In order to solve this problem, a parse tree is needed to understand that \u00e2\u0080\u009cLotus\u00e2\u0080\u009d is not the object of \u00e2\u0080\u009cestimates\u00e2\u0080\u009d.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"There is no relation between these two interpretations other then the fact that their surface forms coincide, and we argue that the only reason to prefer one analysis over the other is compositional.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Sentence (2) and (3) help to disambiguate one way or the other.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"Since different judges judged different systems (recall that judges were excluded to judge system output from their own institution), we normalized the scores.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"This process is repeated 5 times by rotating the data appropriately.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"Overview of the method 2.2 Step by Step Algorithm.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"Recent results (e.g., (Yarowsky 95; Brill 95; Blum and Mitchell 98)) have suggested that unlabeled data can be used quite profitably in reducing the need for supervision.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"The graph was constructed using 2 million trigrams; we chose these by truncating the parallel datasets up to the number of sentence pairs that contained 2 million trigrams.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"A similar maximumlikelihood approach was used by Foster and Kuhn (2007), but for language models only.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"The result of this is shown in Figure 7.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"In order to evaluate and advance this approach, it helps to feed into the knowledge base data that is already enriched with some of the desired information \u00e2\u0080\u0094 as in PCC.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Section 3 discusses the applications that have been completed with PCC, or are under way, or are planned for the future.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"Se ma nti c (a) filters candidate if its semantic tags d o n \u00e2\u0080\u0099 t i n t e r s e c t w i t h t h o s e o f t h e a n a p h o r .","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"Finally, we would like to note that it is possible to devise similar algorithms based with other objective functions than the one given in Equ.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"This task measures how well each package performs in machine translation.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"The method reported in this paper makes use solely of unigram probabilities, and is therefore a zeroeth-order model: the cost of a particular segmentation is estimated as the sum of the costs of the individual words in the segmentation.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"The size of TRIE is particularly sensitive to F1092 c11, so vocabulary filtering is quite effective at reducing model size.","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"More details on the memory-based prediction can be found in Nivre et al. (2004) and Nivre and Scholz (2004).","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"For example, in predicting if a word belongs to a word class, is either true or false, and refers to the surrounding context: if = true, previous word = the otherwise The parameters are estimated by a procedure called Generalized Iterative Scaling (GIS) (Darroch and Ratcliff, 1972).","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"First, it directly encodes linguistic intuitions about POS tag assignments: the model structure reflects the one-tag-per-word property, and a type- level tag prior captures the skew on tag assignments (e.g., there are fewer unique determiners than unique nouns).","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"The PROBING model is 2.4 times as fast as the fastest alternative, SRILM, and uses less memory too.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"Automatic Paraphrase Discovery based on Context and Keywords between NE Pairs","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"Zt can be written as follows Following the derivation of Schapire and Singer, providing that W+ > W_, Equ.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"In this paper, we make a simplifying assumption of one-tag-per-word.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Pairwise comparison is done using the sign test.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"As can be seen in Figure 3, the phrases in the \u00e2\u0080\u009cagree\u00e2\u0080\u009d set include completely different relationships, which are not paraphrases.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"In the second and third translation examples, the IbmS word reordering performs worse than the QmS word reordering, since it can not take properly into account the word reordering due to the German verbgroup.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"The maximum precision row is the upper bound on accuracy if we could pick exactly the correct constituents from among the constituents suggested by the three parsers.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"The evidence may not say anything about whether A is more likely than B, only that C and D are not likely.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"To establish a soft correspondence between the two languages, we use a second similarity function, which leverages standard unsupervised word alignment statistics (\u00a73.3).3 Since we have no labeled foreign data, our goal is to project syntactic information from the English side to the foreign side.","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"The wounds are still healing.), entity-attribute (e.g., She 2001), who determined that in their corpus of German computer tests, 38% of relations were lexically signalled.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Various verbal (e.g., \u00ef\u00bf\u00bd, .::) and adjectival.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"21 84.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"86 78.","label":0},{"source_text":"It outperformed strong unsupervised baselines as well as approaches that relied on direct projections, and bridged the gap between purely supervised and unsupervised POS tagging models.","target_text":"Taking the intersection of languages in these resources, and selecting languages with large amounts of parallel data, yields the following set of eight Indo-European languages: Danish, Dutch, German, Greek, Italian, Portuguese, Spanish and Swedish.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"Unlike the WSJ corpus which has a high frequency of rules like VP \u00e2\u0086\u0092VB PP, Arabic verb phrases usually have lexi calized intervening nodes (e.g., NP subjects and direct objects).","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Following Sproat and Shih (1990), performance for Chinese segmentation systems is generally reported in terms of the dual measures of precision and recalP It is fairly standard to report precision and recall scores in the mid to high 90% range.","label":0},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"We loosely describe the class of all such systems as Linear Context-Free Rewriting Formalisms.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Some of these approaches (e.g., Lin, Chiang, and Su [1993]) attempt to identify unknown words, but do not ac\u00c2\u00ad tually tag the words as belonging to one or another class of expression.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"However, it is desirable if we can separate them.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"The effect of the pruning threshold t0 is shown in Table 5.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"All systems (except for Systran, which was not tuned to Europarl) did considerably worse on outof-domain training data.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"Obviously, the presence of a title after a potential name N increases the probability that N is in fact a name.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"In both cases, the instanceweighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline, and gains of between 0.6 and 1.8 over an equivalent mixture model (with an identical training procedure but without instance weighting).","label":1},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"A modified language model probability p\u00c3\u0086(eje0; e00) is defined as follows: p\u00c3\u0086(eje0; e00) =  1:0 if \u00c3\u0086 = 0 p(eje0; e00) if \u00c3\u0086 = 1 : We associate a distribution p(\u00c3\u0086) with the two cases \u00c3\u0086 = 0 and \u00c3\u0086 = 1 and set p(\u00c3\u0086 = 1) = 0:7.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"In addition to using the simple features directly, we also trained an SVM classifier with these features to distinguish between IN and OUT phrase pairs.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"Each xii is a member of X, where X is a set of possible features.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"4.2 Global Features.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"Pr(eI 1) is the language model of the target language, whereas Pr(fJ 1 jeI1) is the transla tion model.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Indeed, as we shall show in Section 5, even human judges differ when presented with the task of segmenting a text into words, so a definition of the criteria used to determine that a given segmentation is correct is crucial before one can interpret such measures.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"Graph construction for structured prediction problems such as POS tagging is non-trivial: on the one hand, using individual words as the vertices throws away the context necessary for disambiguation; on the other hand, it is unclear how to define (sequence) similarity if the vertices correspond to entire sentences.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"Table 6: Example Translations for the Verbmobil task.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"The index in this array is the vocabulary identifier.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"We utilized two kinds of datasets in our experiments: (i) monolingual treebanks9 and (ii) large amounts of parallel text with English on one side.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"att.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Let say, if we find one system doing better on 20 of the blocks, and worse on 80 of the blocks, is it significantly worse?","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"The supervised POS tagging accuracies (on this tagset) are shown in the last row of Table 2.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"Also, \u00e2\u0080\u009cagree\u00e2\u0080\u009d in the CC-domain is not a desirable keyword.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"However, lexically similar NPs usually refer to the same entity in two cases: proper names and existential noun phrases.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Cur In order to ground such approaches in linguistic observation and description, a multi-level anno 10 For an exposition of the idea as applied to the task of text planning, see (Chiarcos, Stede 2004).","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"We consider two variants of Berg-Kirkpatrick et al.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"IRST is not threadsafe.","label":0},{"source_text":"In order to handle the necessary word reordering as an optimization problem within the dynamic programming approach, they describe a solution to the traveling salesman problem (TSP) which is based on dynamic programming.","target_text":"The computing time is given in terms of CPU time per sentence (on a 450MHz PentiumIIIPC).","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"To set \u03b2, we used the same criterion as for \u03b1, over a dev corpus: The MAP combination was used for TM probabilities only, in part due to a technical difficulty in formulating coherent counts when using standard LM smoothing techniques (Kneser and Ney, 1995).3 Motivated by information retrieval, a number of approaches choose \u201crelevant\u201d sentence pairs from OUT by matching individual source sentences from IN (Hildebrand et al., 2005; L\u00a8u et al., 2007), or individual target hypotheses (Zhao et al., 2004).","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"The parser builds dependency graphs by traversing the input from left to right, using a stack to store tokens that are not yet complete with respect to their dependents.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"Because these two words have identical complements, syntax rules are typically unhelpful for distinguishing between them.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"For example, if Barry is not in commonWords and is found in the list of person first names, then the feature PersonFirstName will be set to 1.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"We then gather all phrases with the same keyword.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"Figure 1 depicts the lattice for a 2-words sentence bclm hneim.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"We are currently working on a complete open source implementation of a training and decoding system, which should become available over the summer. pus, from which also the in-domain test set is taken.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Step 3.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"The ATB annotation distinguishes between verbal and nominal readings of maSdar process nominals.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"The corpus has been annotated with six different types of information, which are characterized in the following subsections.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Since foreign names can be of any length, and since their original pronunciation is effectively unlimited, the identi\u00c2\u00ad fication of such names is tricky.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"Feature weights were set using Och\u2019s MERT algorithm (Och, 2003).","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"However, there will remain a large number of words that are not readily adduced to any produc\u00c2\u00ad tive pattern and that would simply have to be added to the dictionary.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"Because we don\u2019t have a separate development set, we used the training set to select among them and found 0.2 to work slightly better than 0.1 and 0.3.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"evaluated to account for the same fraction of the data.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"Conditioned on T , features of word types W are drawn.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"Thus at each iteration the algorithm is forced to pick features for the location, person and organization in turn for the classifier being trained.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"(S2) The burglar fired the gun three times and fled.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"For every sequence of initial capitalized words, its longest substring that occurs in the same document as a sequence of initCaps is identified.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"The judgement of 4 in the first case will go to a vastly better system output than in the second case.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Next we show that the ATB is similar to other tree- banks in gross statistical terms, but that annotation consistency remains low relative to English (\u00c2\u00a73).","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"The errors shown are from the Berkeley parser output, but they are representative of the other two parsing models.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"Sie.","label":0},{"source_text":"Here both parametric and non-parametric models are explored.","target_text":"As seen by the drop in average individual parser performance baseline, the introduced parser does not perform very well.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Sorted arrays store key-value pairs in an array sorted by key, incurring no space overhead.","label":0},{"source_text":"There is no global pruning.","target_text":"In many cases, there is an even stronger restriction: over large portions of the source string, the alignment is monotone.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"In addition to the Europarl test set, we also collected 29 editorials from the Project Syndicate website2, which are published in all the four languages of the shared task.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"The ratio of buckets to entries is controlled by space multiplier m > 1.","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"This extends previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and using a simpler training procedure.","label":1},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Since the inclusion of out-ofdomain test data was a very late decision, the participants were not informed of this.","label":0},{"source_text":"This corpus has several advantages: it is annotated at different levels.","target_text":"In an experiment on automatic rhetorical parsing, the RST-annotations and PoS tags were used by (Reitter 2003) as a training corpus for statistical classification with Support Vector Machines.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"Adequacy (rank) Fluency (rank) BLEU (rank) upc-jmc (1-7) (1-8) (1-6) lcc (1-6) (1-7) (1-4) utd (1-7) (1-6) (2-7) upc-mr (1-8) (1-6) (1-7) nrc (1-7) (2-6) (8) ntt (1-8) (2-8) (1-7) cmu (3-7) (4-8) (2-7) rali (5-8) (3-9) (3-7) systran (9) (8-9) (10) upv (10) (10) (9) Spanish-English (In Domain) Adequacy (rank) Fluency (rank) BLEU (rank) upc-jmc (1-7) (1-6) (1-5) ntt (1-7) (1-8) (1-5) lcc (1-8) (2-8) (1-4) utd (1-8) (2-7) (1-5) nrc (2-8) (1-9) (6) upc-mr (1-8) (1-6) (7) uedin-birch (1-8) (2-10) (8) rali (3-9) (3-9) (2-5) upc-jg (7-9) (6-9) (9) upv (10) (9-10) (10) German-English (In Domain) Adequacy (rank) Fluency (rank) BLEU (rank) uedin-phi (1-2) (1) (1) lcc (2-7) (2-7) (2) nrc (2-7) (2-6) (5-7) utd (3-7) (2-8) (3-4) ntt (2-9) (2-8) (3-4) upc-mr (3-9) (6-9) (8) rali (4-9) (3-9) (5-7) upc-jmc (2-9) (3-9) (5-7) systran (3-9) (3-9) (10) upv (10) (10) (9) Figure 7: Evaluation of translation to English on in-domain test data 112 English-French (In Domain) Adequacy (rank) Fluency (rank) BLEU (rank) nrc (1-5) (1-5) (1-6) upc-mr (1-4) (1-5) (1-6) upc-jmc (1-6) (1-6) (1-5) systran (2-7) (1-6) (7) utd (3-7) (3-7) (3-6) rali (1-7) (2-7) (1-6) ntt (4-7) (4-7) (1-5) English-Spanish (In Domain) Adequacy (rank) Fluency (rank) BLEU (rank) ms (1-5) (1-7) (7-8) upc-mr (1-4) (1-5) (1-4) utd (1-5) (1-6) (1-4) nrc (2-7) (1-6) (5-6) ntt (3-7) (1-6) (1-4) upc-jmc (2-7) (2-7) (1-4) rali (5-8) (6-8) (5-6) uedin-birch (6-9) (6-10) (7-8) upc-jg (9) (8-10) (9) upv (9-10) (8-10) (10) English-German (In Domain) Adequacy (rank) Fluency (rank) BLEU (rank) upc-mr (1-3) (1-5) (3-5) ntt (1-5) (2-6) (1-3) upc-jmc (1-5) (1-4) (1-3) nrc (2-4) (1-5) (4-5) rali (3-6) (2-6) (1-4) systran (5-6) (3-6) (7) upv (7) (7) (6) Figure 8: Evaluation of translation from English on in-domain test data 113 French-English (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) upc-jmc (1-5) (1-8) (1-4) cmu (1-8) (1-9) (4-7) systran (1-8) (1-7) (9) lcc (1-9) (1-9) (1-5) upc-mr (2-8) (1-7) (1-3) utd (1-9) (1-8) (3-7) ntt (3-9) (1-9) (3-7) nrc (3-8) (3-9) (3-7) rali (4-9) (5-9) (8) upv (10) (10) (10) Spanish-English (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) upc-jmc (1-2) (1-6) (1-3) uedin-birch (1-7) (1-6) (5-8) nrc (2-8) (1-8) (5-7) ntt (2-7) (2-6) (3-4) upc-mr (2-8) (1-7) (5-8) lcc (4-9) (3-7) (1-4) utd (2-9) (2-8) (1-3) upc-jg (4-9) (7-9) (9) rali (4-9) (6-9) (6-8) upv (10) (10) (10) German-English (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) systran (1-4) (1-4) (7-9) uedin-phi (1-6) (1-7) (1) lcc (1-6) (1-7) (2-3) utd (2-7) (2-6) (4-6) ntt (1-9) (1-7) (3-5) nrc (3-8) (2-8) (7-8) upc-mr (4-8) (6-8) (4-6) upc-jmc (4-8) (3-9) (2-5) rali (8-9) (8-9) (8-9) upv (10) (10) (10) Figure 9: Evaluation of translation to English on out-of-domain test data 114 English-French (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) systran (1) (1) (1) upc-jmc (2-5) (2-4) (2-6) upc-mr (2-4) (2-4) (2-6) utd (2-6) (2-6) (7) rali (4-7) (5-7) (2-6) nrc (4-7) (4-7) (2-5) ntt (4-7) (4-7) (3-6) English-Spanish (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) upc-mr (1-3) (1-6) (1-2) ms (1-7) (1-8) (6-7) utd (2-6) (1-7) (3-5) nrc (1-6) (2-7) (3-5) upc-jmc (2-7) (1-6) (3-5) ntt (2-7) (1-7) (1-2) rali (6-8) (4-8) (6-8) uedin-birch (6-10) (5-9) (7-8) upc-jg (8-9) (9-10) (9) upv (9) (8-9) (10) English-German (Out of Domain) Adequacy (rank) Fluency (rank) BLEU (rank) systran (1) (1-2) (1-6) upc-mr (2-3) (1-3) (1-5) upc-jmc (2-3) (3-6) (1-6) rali (4-6) (4-6) (1-6) nrc (4-6) (2-6) (2-6) ntt (4-6) (3-5) (1-6) upv (7) (7) (7) Figure 10: Evaluation of translation from English on out-of-domain test data 115 French-English In domain Out of Domain Adequacy Adequacy 0.3 0.3 \u2022 0.2 0.2 0.1 0.1 -0.0 -0.0 -0.1 -0.1 -0.2 -0.2 -0.3 -0.3 -0.4 -0.4 -0.5 -0.5 -0.6 -0.6 -0.7 -0.7 \u2022upv -0.8 -0.8 21 22 23 24 25 26 27 28 29 30 31 15 16 17 18 19 20 21 22 \u2022upv \u2022systran upcntt \u2022 rali upc-jmc \u2022 cc Fluency 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 \u2022upv -0.5 \u2022systran \u2022upv upc -jmc \u2022 Fluency 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 \u2022 \u2022 \u2022 td t cc upc- \u2022 rali 21 22 23 24 25 26 27 28 29 30 31 15 16 17 18 19 20 21 22 Figure 11: Correlation between manual and automatic scores for French-English 116 Spanish-English Figure 12: Correlation between manual and automatic scores for Spanish-English -0.3 -0.4 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 \u2022upv -0.4 \u2022upv -0.3 In Domain \u2022upc-jg Adequacy 0.3 0.2 0.1 -0.0 -0.1 -0.2 Out of Domain \u2022upc-jmc \u2022nrc \u2022ntt Adequacy upc-jmc \u2022 \u2022 \u2022lcc \u2022 rali \u2022 \u2022rali -0.7 -0.5 -0.6 \u2022upv 23 24 25 26 27 28 29 30 31 32 19 20 21 22 23 24 25 26 27 28 \u2022 \u2022rali Fluency 0.2 0.1 -0.0 -0.1 -0.2 ntt \u2022 upc-mr \u2022lcc \u2022utd \u2022upc-jg \u2022rali Fluency 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 \u2022upc-jmc \u2022 uedin-birch -0.5 -0.5 \u2022upv 23 24 25 26 27 28 29 30 31 32 19 20 21 22 23 24 25 26 27 28 117 In Domain Out of Domain Adequacy Adequacy German-English 15 16 17 18 19 20 21 22 23 24 25 26 27 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 lcc \u2022 upc-jmc \u2022systran \u2022upv Fluency \u2022ula \u2022upc-mr \u2022lcc 15 16 17 18 19 20 21 22 23 24 25 26 27 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 \u2022systran \u2022upv \u2022uedin-phi -jmc \u2022rali \u2022systran -0.3 -0.4 -0.5 -0.6 \u2022upv 12 13 14 15 16 17 18 19 20 0.4 0.3 0.2 0.1 -0.0 -0.1 -0.2 Fluency uedin-phi \u2022 \u2022 \u2022utd \u2022upc-jmc \u2022upc-mr 0.4 \u2022rali -0.3 -0.4 -0.5 \u2022upv 12 13 14 15 16 17 18 19 20 0.3 0.2 0.1 -0.0 -0.1 -0.2 English-French In Domain Out of Domain Adequacy Adequacy .","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"It is important to bear in mind, though, that this is not an inherent limitation of the model.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"Note that on some examples (around 2% of the test set) CoBoost abstained altogether; in these cases we labeled the test example with the baseline, organization, label.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Human evaluation is one way to distinguish between the two cases.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"D o m ai n Li n k ac cu ra cy W N c o v e r a g e C C 7 3 . 3 % 2 \/ 1 1 P C 8 8 . 9 % 2 \/ 8 Table 2.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"Tag set As is standard, for all experiments, we set the number of latent model tag states to the size of the annotated tag set.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"We asked six native speakers-three from Taiwan (TlT3), and three from the Mainland (M1M3)-to segment the corpus.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"Other approaches encode sparsity as a soft constraint.","label":0},{"source_text":"They proposed an unsupervised method to discover paraphrases from a large untagged corpus.","target_text":"At this step, we will try to link those sets, and put them into a single cluster.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"For even larger models, we recommend RandLM; the memory consumption of the cache is not expected to grow with model size, and it has been reported to scale well.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"2 61.7 64.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"Applications such as machine translation use language model probability as a feature to assist in choosing between hypotheses.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"As we have seen, the lexicon of basic words and stems is represented as a WFST; most arcs in this WFST represent mappings between hanzi and pronunciations, and are costless.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Performance improvements transfer to the Moses (Koehn et al., 2007), cdec (Dyer et al., 2010), and Joshua (Li et al., 2009) translation systems where our code has been integrated.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Table 8b shows that verbal nouns are the hardest pre-terminal categories to identify.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"(We would like to note though that unlike previous boosting algorithms, the CoBoost algorithm presented here is not a boosting algorithm under Valiant's (Valiant 84) Probably Approximately Correct (PAC) model.)","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"In the case of adverbial reduplication illustrated in (3b) an adjective of the form AB is reduplicated as AABB.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"The domain is general politics, economics and science.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"A dynamic programming algorithm is then used to select the sequence of word classes with the highest probability.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"The taggers were trained on datasets labeled with the universal tags.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"The second row is the accuracy of the best of the three parsers.'","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"For example, it is well-known that one can build a finite-state bigram (word) model by simply assigning a state Si to each word Wi in the vocabulary, and having (word) arcs leaving that state weighted such that for each Wj and corresponding arc aj leaving Si, the cost on aj is the bigram cost of WiWj- (Costs for unseen bigrams in such a scheme would typically be modeled with a special backoff state.)","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"At each point during the derivation, the parser has a choice between pushing the next input token onto the stack \u2013 with or without adding an arc from the token on top of the stack to the token pushed \u2013 and popping a token from the stack \u2013 with or without adding an arc from the next input token to the token popped.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"Past work however, has typically associ n = n P (Ti)P (Wi|Ti) = i=1 1 n K n ated these features with token occurrences, typically in an HMM.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"The details are given in (Och and Ney, 2000).","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"We focus here instead on adapting the two most important features: the language model (LM), which estimates the probability p(wIh) of a target word w following an ngram h; and the translation models (TM) p(slt) and p(t1s), which give the probability of source phrase s translating to target phrase t, and vice versa.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seemed to be less pronounced on out-of-domain test data.","target_text":"We computed BLEU scores for each submission with a single reference translation.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"We measured recall (Rec), precision (Pr), and the F-measure (F) with recall and precision equally weighted.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"Even without features, but still using the tag prior, our median result is 52.0%, still significantly outperforming Grac\u00c2\u00b8a et al.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Following Sproat and Shih (1990), performance for Chinese segmentation systems is generally reported in terms of the dual measures of precision and recalP It is fairly standard to report precision and recall scores in the mid to high 90% range.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"This group contains a large number of features (one for each token string present in the training data).","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"This is manifest in the lexical choices but 1 www.coli.unisb.de\/\u00e2\u0088\u00bcthorsten\/tnt\/ Dagmar Ziegler is up to her neck in debt.","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"There is a \u00e2\u0080\u0098core corpus\u00e2\u0080\u0099 of ten commentaries, for which the range of information (except for syntax) has been completed; the remaining data has been annotated to different degrees, as explained below.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"An important subproblem of language model storage is therefore sparse mapping: storing values for sparse keys using little memory then retrieving values given keys using little time.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"01 75.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"We first define &quot;pseudo-labels&quot;,-yt, as follows: = Yi t sign(g 0\\ 2\u2014 kx2,m < i < n Thus the first m labels are simply copied from the labeled examples, while the remaining (n \u2014 m) examples are taken as the current output of the second classifier.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"We are unaware of prior results for the Stanford parser.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"Across all languages, +PRIOR consistently outperforms 1TW, reducing error on average by 9.1% and 5.9% on best and median settings respectively.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Mikheev et al.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"If the parse contains productions from outside our grammar the machine has no direct method for handling them (e.g. the resulting database query may be syntactically malformed).","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"MENE is then trained on 80% of the training corpus, and tested on the remaining 20%.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"The kind of annotation work presented here would clearly benefit from the emergence of standard formats and tag sets, which could lead to sharable resources of larger size.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"Two subjects are each given a calendar and they are asked to schedule a meeting.","label":0},{"source_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","target_text":"Additionally, for a given coverage set, at most 250 different hypotheses are kept during the search process, and the number of different words to be hypothesized by a source word is limited.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"The similar explanation applies to the link to the \u00e2\u0080\u009cstake\u00e2\u0080\u009d set.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"Previous approaches have tried to find examples that are similar to the target domain.","label":0},{"source_text":"Here we present two algorithms.","target_text":"(3)), with one term for each classifier.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"With some minor API changes, namely returning the length of the n-gram matched, it could also be faster\u2014though this would be at the expense of an optimization we explain in Section 4.1.","label":0},{"source_text":"Here we present two algorithms.","target_text":".","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"The other half was replaced by other participants, so we ended up with roughly the same number.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"We incur some additional memory cost due to storing state in each hypothesis, though this is minimal compared with the size of the model itself.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"This is especially bad with PROBING because it is based on hashing and performs random lookups, but it is not intended to be used in low-memory scenarios.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"In addition to the named-entity string (Maury Cooper or Georgia), a contextual predictor was also extracted.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"For each set, the phrases with bracketed frequencies are considered not paraphrases in the set.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"This design does not guarantee \u00e2\u0080\u009cstructural zeros,\u00e2\u0080\u009d but biases towards sparsity.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"We follow the guidelines developed in the TIGER project (Brants et al. 2002) for syntactic annotation of German newspaper text, using the Annotate3 tool for interactive construction of tree structures.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"The toplevel weights are trained to maximize a metric such as BLEU on a small development set of approximately 1000 sentence pairs.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"The BLEU score has been shown to correlate well with human judgement, when statistical machine translation systems are compared (Doddington, 2002; Przybocki, 2004; Li, 2005).","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"The errors shown are from the Berkeley parser output, but they are representative of the other two parsing models.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"Recent work has made significant progress on unsupervised POS tagging (Me\u00c2\u00b4rialdo, 1994; Smith and Eisner, 2005; Haghighi and Klein, 2006; Johnson,2007; Goldwater and Griffiths, 2007; Gao and John son, 2008; Ravi and Knight, 2009).","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"We experiment with four values for each hyperparameter resulting in 16 (\u00ce\u00b1, \u00ce\u00b2) combinations: \u00ce\u00b1 \u00ce\u00b2 0.001, 0.01, 0.1, 1.0 0.01, 0.1, 1.0, 10 Iterations In each run, we performed 30 iterations of Gibbs sampling for the type assignment variables W .4 We use the final sample for evaluation.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"For each candidate antecedent, BABAR identifies the caseframe that would extract the candidate, pairs it with the anaphor\u00e2\u0080\u0099s caseframe, and consults the CF Network to see if this pair of caseframes has co-occurred in previous resolutions.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"The most frequently occurring last words of organization names in cslist are compiled into a list of corporate suffixes, Corporate-Suffix- List.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"The global feature groups are: InitCaps of Other Occurrences (ICOC): There are 2 features in this group, checking for whether the first occurrence of the same word in an unambiguous position (non first-words in the TXT or TEXT zones) in the same document is initCaps or not-initCaps.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"\u00e2\u0080\u009cThe gun\u00e2\u0080\u009d will be extracted by the caseframe \u00e2\u0080\u009cfired <patient>\u00e2\u0080\u009d.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"We used it to score all phrase pairs in the OUT table, in order to provide a feature for the instance-weighting model.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"An extended lexicon model is defined, and its likelihood is compared to a baseline lexicon model, which takes only single-word dependencies into account.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"But we also need an estimate of the probability for a non-occurring though possible plural form like i\u00c2\u00a5JJ1l.f, nan2gua1-men0 'pumpkins.'","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"In any event, to date, we have not compared different methods for deriving the set of initial frequency estimates.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"The structure uses linear probing hash tables and is designed for speed.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"The substructures that are unanimously hypothesized by the parsers should be preserved after combination, and the combination technique should not foolishly create substructures for which there is no supporting evidence.","label":0},{"source_text":"Their results suggested that it was possible to learn accurate POS taggers for languages which did not have any annotated data, but have translations into a resource-rich language.","target_text":"Because there might be some controversy about the exact definitions of such universals, this set of coarse-grained POS categories is defined operationally, by collapsing language (or treebank) specific distinctions to a set of categories that exists across all languages.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"Denote by g3(x) = Et crithl(x) , j E {1,2} the unthresholded strong-hypothesis (i.e., f3 (x) = sign(gi (x))).","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"na me =>1 ha nzi fa mi ly 1 ha nzi gi ve n 4.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"wo rd => na m e 2.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"We will refer to the semantic classes that co-occur with a caseframe as the semantic expectations of the caseframe.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"We can use the preceding linguistic and annotation insights to build a manually annotated Arabic grammar in the manner of Klein and Manning (2003).","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"In the ATB, :: b asta\u00e2\u0080\u0099adah is tagged 48 times as a noun and 9 times as verbal noun.","label":0},{"source_text":"In this paper, the authors are of the opinion that the sequence models-based approaches usually treat token-level tag assignment as the primary latent variable.","target_text":"Second, the reduced number of hidden variables and parameters dramatically speeds up learning and inference.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"The matching sentence pairs are then added to the IN corpus, and the system is re-trained.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"Also, the argument has been made that machine translation performance should be evaluated via task-based evaluation metrics, i.e. how much it assists performing a useful task, such as supporting human translators or aiding the analysis of texts.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"The contextual role knowledge had the greatest impact on pronouns: +13% recall for terrorism and +15% recall for disasters, with a +1% precision gain in terrorism and a small precision drop of -3% in disasters.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"Compared to the EMEA\/EP setting, the two domains in the NIST setting are less homogeneous and more similar to each other; there is also considerably more IN text available.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Another way to view the judgements is that they are less quality judgements of machine translation systems per se, but rankings of machine translation systems.","label":0},{"source_text":"Here we show how non-projective dependency parsing can be achieved by combining a data driven projective parser with special graph transformation techniques.","target_text":"In the following, we use the notation wi wj to mean that (wi, r, wj) E A; r we also use wi wj to denote an arc with unspecified label and wi \u2014*\u2217 wj for the reflexive and transitive closure of the (unlabeled) arc relation.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"During training, we treat as observed the language word types W as well as the token-level corpus w. We utilize Gibbs sampling to approximate our collapsed model posterior: P (T ,t|W , w, \u00ce\u00b1, \u00ce\u00b2) \u00e2\u0088\u009d P (T , t, W , w|\u00ce\u00b1, \u00ce\u00b2) 0.7 0.6 0.5 0.4 0.3 English Danish Dutch Germany Portuguese Spanish Swedish = P (T , t, W , w, \u00cf\u0088, \u00ce\u00b8, \u00cf\u0086, w|\u00ce\u00b1, \u00ce\u00b2)d\u00cf\u0088d\u00ce\u00b8d\u00cf\u0086 Note that given tag assignments T , there is only one setting of token-level tags t which has mass in the above posterior.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"In the \u00e2\u0080\u009cPerson \u00e2\u0080\u0093 Person\u00e2\u0080\u009d domain, 618 keywords are found, and in the \u00e2\u0080\u009cCountry \u00e2\u0080\u0093 Country\u00e2\u0080\u009d domain, 303 keywords are found.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"The method halves the error rate in comparison to a method using the labeled examples alone.","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"Thus, most broad-coverage parsers based on dependency grammar have been restricted to projective structures.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"Roughly speaking, previous work can be divided into three categories, namely purely statistical approaches, purely lexi\u00c2\u00ad cal rule-based approaches, and approaches that combine lexical information with sta\u00c2\u00ad tistical information.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"An edge indicates that the two features must have the same label.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"In this paper, we describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"On several languages, we report performance exceeding that of state-of-the art systems.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"It is worth noting that, although nonprojective constructions are less frequent in DDT than in PDT, they seem to be more deeply nested, since only about 80% can be projectivized with a single lift, while almost 95% of the non-projective arcs in PDT only require a single lift.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"1 61.7 37.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"For example, we may have semantic information (e.g. database query operations) associated with the productions in a grammar.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"To our knowledge, ours is the first analysis of this kind for Arabic parsing.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"But the city name Sharm Al- Sheikh is also iDafa, hence the possibility for the incorrect annotation in (b).","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"2.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"Note that in line 4 the last visited position for the successor hypothesis must be m. Otherwise , there will be four uncovered positions for the predecessor hypothesis violating the restriction.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"In Equations 1 through 3 we develop the model for constructing our parse using na\u00efve Bayes classification.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"Aside from adding a simple rule to correct alif deletion caused by the preposition J, no other language-specific processing is performed.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"The way we cant distinction between system performance.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"The corpus was wordaligned using both HMM and IBM2 models, and the phrase table was the union of phrases extracted from these separate alignments, with a length limit of 7.","label":0},{"source_text":"The authors believe automatic paraphrase discovery is an important component for building a fully automatic information extraction system.","target_text":"Step 3.","label":0},{"source_text":"Instance-weighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline.","target_text":"For the TM, this is: where cI(s, t) is the count in the IN phrase table of pair (s, t), po(s|t) is its probability under the OUT TM, and cI(t) = &quot;s, cI(s', t).","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"The model was parameterized such that the joint probability of a (label, feature-set) pair P(yi, xi) is written as The model assumes that (y, x) pairs are generated by an underlying process where the label is first chosen with some prior probability P(yi); the number of features mi is then chosen with some probability P(mi); finally the features are independently generated with probabilities P(xulyi).","label":0},{"source_text":"In order to create good-sized vectors for similarity calculation, they had to set a high frequency threshold.","target_text":"Step 4.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Though we do not directly compare state implementations, performance metrics in Table 1 indicate our overall method is faster.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"The translation scores for the hypotheses generated with different threshold values t0 are compared to the translation scores obtained with a conservatively large threshold t0 = 10:0 . For each test series, we count the number of sentences whose score is worse than the corresponding score of the test series with the conservatively large threshold t0 = 10:0, and this number is reported as the number of search errors.","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"A graph D = (W, A) is well-formed iff it is acyclic and connected.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"For example, suppose the current model assigns a belief value of .60 to {A, B}, meaning that it is 60% sure that the correct hypothesis is either A or B. Then new evidence arrives with a belief value of .70 assigned 5 Initially there are no competing hypotheses because all hypotheses are included in \u00ce\u00b8 by definition.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"The way we cant distinction between system performance.","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"In section 2 we introduce the graph transformation techniques used to projectivize and deprojectivize dependency graphs, and in section 3 we describe the data-driven dependency parser that is the core of our system.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"We use v1.0 mainly because previous studies on joint inference reported results w.r.t. v1.0 only.5 We expect that using the same setup on v2.0 will allow a crosstreebank comparison.6 We used the first 500 sentences as our dev set and the rest 4500 for training and report our main results on this split.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"C is the union of the sets of constituents suggested by the parsers. r(c) is a binary function returning t (for true) precisely when the constituent c E C should be included in the hypothesis.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"The TRIE model continues to use the least memory of ing (-P) with MAP POPULATE, the default.","label":0},{"source_text":"This paper talks about Pseudo-Projective Dependency Parsing.","target_text":"The Prague Dependency Treebank (PDT) consists of more than 1M words of newspaper text, annotated on three levels, the morphological, analytical and tectogrammatical levels (Haji\u02c7c, 1998).","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"This intuition is born out by the experimental results.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"Each learner is free to pick the labels for these instances.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"We also thank ChaoHuang Chang, reviewers for the 1994 ACL conference, and four anonymous reviewers for Computational Linguistics for useful comments.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"We are currently working on a complete open source implementation of a training and decoding system, which should become available over the summer. pus, from which also the in-domain test set is taken.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"Lossy compressed models RandLM (Talbot and Osborne, 2007) and Sheffield (Guthrie and Hepple, 2010) offer better memory consumption at the expense of CPU and accuracy.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"No: Predecessor coverage set Successor coverage set 1 (f1; ;mg n flg ; l0) !","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"Since our graph is built from a parallel corpus, we can use standard word alignment techniques to align the English sentences De 5Note that many combinations are impossible giving a PMI value of 0; e.g., when the trigram type and the feature instantiation don\u2019t have words in common. and their foreign language translations Df.6 Label propagation in the graph will provide coverage and high recall, and we therefore extract only intersected high-confidence (> 0.9) alignments De\ufffdf.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"An initial step of any text\u00c2\u00ad analysis task is the tokenization of the input into words.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"63 81.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"Also, \u00e2\u0080\u009cagree\u00e2\u0080\u009d in the CC-domain is not a desirable keyword.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"This approach leads to a search procedure with complexity O(E3 J4).","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"Both of the switching techniques, as well as the parametric hybridization technique were also shown to be robust when a poor parser was introduced into the experiments.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"Our baseline for all sentence lengths is 5.23% F1 higher than the best previous result.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"We develop our POS induction model based on the feature-based HMM of Berg-Kirkpatrick et al. (2010).","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"This is especially effective at reducing load time, since raw bytes are read directly to memory\u2014or, as happens with repeatedly used models, are already in the disk cache.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"By sharing stacks (in IG's) or by using nonlinear equations over f-structures (in FUG's and LFG's), structures with unbounded dependencies between paths can be generated.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"For definite NPs, the results are a mixed bag: some knowledge sources increased recall a little, but at the expense of some precision.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"Statistical NERs usually find the sequence of tags that maximizes the probability , where is the sequence of words in a sentence, and is the sequence of named-entity tags assigned to the words in . Attempts have been made to use global information (e.g., the same named entity occurring in different sentences of the same document), but they usually consist of incorporating an additional classifier, which tries to correct the errors in the output of a first NER (Mikheev et al., 1998; Borthwick, 1999).","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"For that application, at a minimum, one would want to know the phonological word boundaries.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":" We evaluate the system's performance by comparing its segmentation 'Tudgments\" with the judgments of a pool of human segmenters, and the system is shown to perform quite well.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"Limitations of (Blum and Mitchell 98): While the assumptions of (Blum and Mitchell 98) are useful in developing both theoretical results and an intuition for the problem, the assumptions are quite limited.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"Hyperparameters Our model has two Dirichlet concentration hyperparameters: \u00ce\u00b1 is the shared hyperparameter for the token-level HMM emission and transition distributions.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"The corresponding token words w are drawn conditioned on t and \u00ce\u00b8.2 Our full generative model is given by: K P (\u00cf\u0086, \u00ce\u00b8|T , \u00ce\u00b1, \u00ce\u00b2) = n (P (\u00cf\u0086t|\u00ce\u00b1)P (\u00ce\u00b8t|T , \u00ce\u00b1)) t=1 The transition distribution \u00cf\u0086t for each tag t is drawn according to DIRICHLET(\u00ce\u00b1, K ), where \u00ce\u00b1 is the shared transition and emission distribution hyperparameter.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"Thus, provided at least this amount of IN data is available\u2014as it is in our setting\u2014adapting these weights is straightforward.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Sparse lookup is a key subproblem of language model queries.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"We trained this model by optimizing the following objective function: Note that this involves marginalizing out all possible state configurations z for a sentence x, resulting in a non-convex objective.","label":0},{"source_text":"Using less training data than other systems, their NER can perform as well as other state-of-the-art NERs.","target_text":"The system will then look for sequences of initial capitalized words that match the acronyms found in the whole document.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"Our parsing performance measures (SY N) thus report the PARSEVAL extension proposed in Tsarfaty (2006).","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"If one system is perfect, another has slight flaws and the third more flaws, a judge is inclined to hand out judgements of 5, 4, and 3.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"Replacing this with an ranked evaluation seems to be more suitable.","label":1},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"While it is possible to derive a closed form solution for this convex objective function, it would require the inversion of a matrix of order |Vf|.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"As we have noted in Section 2, the general semantic class to which a hanzi belongs is often predictable from its semantic radical.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"We propose a limit of 70 words for Arabic parsing evaluations.","label":0},{"source_text":"This paper offers a broad insight into of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","target_text":"30 75.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"On several languages, we report performance exceeding that of state-of-the art systems.","label":0},{"source_text":"This paper talks about Exploiting Diversity in Natural Language Processing: Combining Parsers.","target_text":"This technique has the advantage of requiring no training, but it has the disadvantage of treating all parsers equally even though they may have differing accuracies or may specialize in modeling different phenomena.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"The role that each noun phrase plays in the kidnapping event is key to distinguishing these cases.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"This paper, however, provides a comprehensive overview of the data collection effort and its current state.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"Each name class is subdivided into 4 sub-classes, i.e., N begin, N continue, N end, and N unique.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Diacritics can also be used to specify grammatical relations such as case and gender.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"We compared the ATB5 to tree- banks for Chinese (CTB6), German (Negra), and English (WSJ) (Table 4).","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Section 3 discusses the applications that have been completed with PCC, or are under way, or are planned for the future.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"The numbers falling into the location, person, organization categories were 186, 289 and 402 respectively.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"While the first three models get three to four tags wrong, our best model gets only one word wrong and is the most accurate among the four models for this example.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"lla\/llb and 14a\/14b respectively).","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"(7), such as the likelihood function used in maximum-entropy problems and other generalized additive models (Lafferty 99).","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"An easy way to achieve this is to put the domain-specific LMs and TMs into the top-level log-linear model and learn optimal weights with MERT (Och, 2003).","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"VBD she added VP PUNC \u00e2\u0080\u009c SBAR IN NP 0 NN.","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"With respect to exact match, the improvement is even more noticeable, which shows quite clearly that even if non-projective dependencies are rare on the token level, they are nevertheless important for getting the global syntactic structure correct.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"The experimental tests are carried out on the Verbmobil task (GermanEnglish, 8000-word vocabulary), which is a limited-domain spoken-language task.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"We used section 23 as the development set for our combining techniques, and section 22 only for final testing.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"For example, one parser could be more accurate at predicting noun phrases than the other parsers.","label":0},{"source_text":"In this paper the author evaluates machine translation performance for six European language pairs that participated in a shared task: translating French, German, Spanish texts to English and back.","target_text":"The manual evaluation of scoring translation on a graded scale from 1\u20135 seems to be very hard to perform.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Compared to decoding, this task is cache-unfriendly in that repeated queries happen only as they naturally occur in text.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"For this example, since the sequence Even News Broadcasting Corp. only appears once in the document, its longest substring that occurs in the same document is News Broadcasting Corp. In this case, News has an additional feature of I begin set to 1, Broadcasting has an additional feature of I continue set to 1, and Corp. has an additional feature of I end set to 1.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"For a given \"word\" in the automatic segmentation, if at least k of the hu\u00c2\u00ad man judges agree that this is a word, then that word is considered to be correct.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"The global feature groups are: InitCaps of Other Occurrences (ICOC): There are 2 features in this group, checking for whether the first occurrence of the same word in an unambiguous position (non first-words in the TXT or TEXT zones) in the same document is initCaps or not-initCaps.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"With the additional assumption that (s, t) can be restricted to the support of co(s, t), this is equivalent to a \u201cflat\u201d alternative to (6) in which each non-zero co(s, t) is set to one.","label":0},{"source_text":"One can trivially create situations in which strictly binary-branching trees are combined to create a tree with only the root node and the terminal nodes, a completely flat structure.","target_text":"In Equations 1 through 3 we develop the model for constructing our parse using na\u00efve Bayes classification.","label":0},{"source_text":"The second algorithm builds on a boosting algorithm called AdaBoost.","target_text":"Schapire and Singer show that the training error is bounded above by Thus, in order to greedily minimize an upper bound on training error, on each iteration we should search for the weak hypothesis ht and the weight at that minimize Z.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"-1 means that an NP should be ruled out as a possible antecedent, and 0 means that the knowledge source remains neutral (i.e., it has no reason to believe that they cannot be coreferent).","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"Microsoft\u2019s approach uses dependency trees, others use hierarchical phrase models.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"These models generally outperform our memory consumption but are much slower, even when cached.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"This revealed interesting clues about the properties of automatic and manual scoring.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"For each sentence, we counted how many n-grams in the system output also occurred in the reference translation.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"For the examples given in (1) and (2) this certainly seems possible.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Figure 1 shows an example.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"We propose maximizing , where is the sequence of named- entity tags assigned to the words in the sentence , and is the information that can be extracted from the whole document containing . Our system is built on a maximum entropy classifier.","label":0},{"source_text":"This paper talks about Unsupervised Models for Named Entity Classification.","target_text":"Unsupervised Models for Named Entity Classification Collins","label":1},{"source_text":"They showed that it was useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties their trees, find that several of the formalisms considered can be seen as being closely related since they have derivation tree sets with the same structure as those produced by Context-Free Grammars.","target_text":"Since these path sets are semilinear, the property of independent paths in their tree sets is sufficient to cause semilinearity of the languages generated by them.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"(See also Wu and Fung [1994].)","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"Due to many similarly performing systems, we are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","label":0},{"source_text":"A large number of rules are needed for coverage of the domain, suggesting that a fairly large number of labeled examples should be required to train a classifier.","target_text":"This section describes AdaBoost, which is the basis for the CoBoost algorithm.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"Like in the co-reference annotation, G\u00c2\u00a8otze\u00e2\u0080\u0099s proposal has been applied by two annotators to the core corpus but it has not been systematically evaluated yet.","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"Each name class is subdivided into 4 sub-classes, i.e., N begin, N continue, N end, and N unique.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"Other approaches encode sparsity as a soft constraint.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"3.1 Maximum Entropy.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"We present a coreference resolver called BABAR that uses contextual role knowledge to evaluate possible antecedents for an anaphor.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"Hence, we use the bootstrap resampling method described by Koehn (2004).","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"For a given \"word\" in the automatic segmentation, if at least k of the hu\u00c2\u00ad man judges agree that this is a word, then that word is considered to be correct.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"A list of words occurring more than 10 times in the training data is also collected (commonWords).","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"This decreases the statistical significance of our results compared to those studies.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"It is formally straightforward to extend the grammar to include these names, though it does increase the likelihood of overgeneration and we are unaware of any working systems that incorporate this type of name.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"First of all, most previous articles report perfor\u00c2\u00ad mance in terms of a single percent-correct score, or else in terms of the paired measures of precision and recall.","label":0},{"source_text":"Here we present two algorithms.","target_text":"For example, in .., says Mr. Cooper, a vice president of.. both a spelling feature (that the string contains Mr.) and a contextual feature (that president modifies the string) are strong indications that Mr. Cooper is of type Person.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"The complexity of the quasimonotone search is O(E3 J (R2+LR)).","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"This is an iterative method that improves the estimation of the parameters at each iteration.","label":0},{"source_text":"Das and Petrov, in this paper, approached inducing unsupervised part-of-speech taggers for languages that had no labeled training data, but had translated text in a resource-rich language.","target_text":"In this particular case, all English vertices are labeled as nouns by the supervised tagger.","label":0},{"source_text":"Foster et all describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","target_text":"It is also worth pointing out a connection with Daum\u00b4e\u2019s (2007) work that splits each feature into domain-specific and general copies.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"Three state-of-the-art statistical parsers are combined to produce more accurate parses, as well as new bounds on achievable Treebank parsing accuracy.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"It is. based on the traditional character set rather than the simplified character set used in Singapore and Mainland China.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"Intuitively, as suggested by the example in the introduction, this is the right granularity to capture domain effects.","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"Among the IS-units, the referring expressions are marked as such and will in the second phase receive a label for cognitive status (active, accessible- text, accessible-situation, inferrable, inactive).","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"In the labeled version of these metrics (L) both heads and arc labels must be correct, while the unlabeled version (U) only considers heads.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"The way we cant distinction between system performance.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"Our approach uses word-to-word dependencies between source and target words.","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"Chinese word segmentation can be viewed as a stochastic transduction problem.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"17 They also provide a set of title-driven rules to identify names when they occur before titles such as $t. 1: xianlshengl 'Mr.' or i:l:itr!J tai2bei3 shi4zhang3 'Taipei Mayor.'","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"This is the first time that we organized a large-scale manual evaluation.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Our results show that MENERGI performs as well as IdentiFinder when trained on comparable amount of training data.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"We report micro-averaged (whole corpus) and macro-averaged (per sentence) scores along add a constraint on the removal of punctuation, which has a single tag (PUNC) in the ATB.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"In this case, this knowledge source reports that the candidate is not a viable antecedent for the anaphor.","label":0},{"source_text":"Each out-of-domain phrase pair was characterized by a set of simple features intended to reflect how useful it would be.","target_text":"However, when the linear LM is combined with a linear TM (lm+lin tm) or MAP TM (lm+map TM), the results are much better than a log-linear combination for the EMEA setting, and on a par for NIST.","label":0},{"source_text":"We chose one of four labels for each example: location, person, organization, or noise where the noise category was used for items that were outside the three categories.","target_text":"Note that in our formalism a weakhypothesis can abstain.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Interestingly, Chang et al. report 80.67% recall and 91.87% precision on an 11,000 word corpus: seemingly, our system finds as many names as their system, but with four times as many false hits.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"In both cases, the instanceweighting approach improved over a wide range of baselines, giving gains of over 2 BLEU points over the best non-adapted baseline, and gains of between 0.6 and 1.8 over an equivalent mixture model (with an identical training procedure but without instance weighting).","label":0},{"source_text":"The evaluation compares the performance of the system with that of several human judges and inter-human agreement on a single correct way to segment a text.","target_text":"The most accurate characterization of Chinese writing is that it is morphosyllabic (DeFrancis 1984): each hanzi represents one morpheme lexically and semantically, and one syllable phonologi\u00c2\u00ad cally.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"This is less effective in our setting, where IN and OUT are disparate.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"A similar structure is used in speech recognition.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Reiche\u00e2\u0080\u0099s colleagues will make sure that the concept is waterproof.","label":0},{"source_text":"The code is open source, has minimal dependencies, and offers both C++ and Java interfaces for integration.","target_text":"Our open-source (LGPL) implementation is also available for download as a standalone package with minimal (POSIX and g++) dependencies.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"We formulate the update as follows: where \u2200ui \u2208 Vf \\ Vfl, \u03b3i(y) and \u03bai are defined as: We ran this procedure for 10 iterations.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"The most accurate characterization of Chinese writing is that it is morphosyllabic (DeFrancis 1984): each hanzi represents one morpheme lexically and semantically, and one syllable phonologi\u00c2\u00ad cally.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"This method, one instance of which we term the \"greedy algorithm\" in our evaluation of our own system in Section 5, involves starting at the beginning (or end) of the sentence, finding the longest word starting (ending) at that point, and then repeating the process starting at the next (previous) hanzi until the end (begin\u00c2\u00ad ning) of the sentence is reached.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"(Blum and Mitchell 98) go on to give PAC results for learning in the cotraining case.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"On one end of the spectrum are clustering approaches that assign a single POS tag to each word type (Schutze, 1995; Lamar et al., 2010).","label":0},{"source_text":"In order to realize the full potential of dependency-based syntactic parsing, it is desirable to allow non-projective dependency structures.","target_text":"More details on the memory-based prediction can be found in Nivre et al. (2004) and Nivre and Scholz (2004).","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"Since these are distinct syntactic units, they are typically segmented.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"To facilitate the comparison of our results to those reported by (Cohen and Smith, 2007) we use their data set in which 177 empty and \u201cmalformed\u201d7 were removed.","label":0},{"source_text":"The use of global features has shown excellent result in the performance on MUC-6 and MUC-7 test data.","target_text":"Each name class is subdivided into 4 sub-classes, i.e., N begin, N continue, N end, and N unique.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"It may be more realistic to replace the second criteria with a softer one, for example (Blum and Mitchell 98) suggest the alternative Alternatively, if Ii and 12 are probabilistic learners, it might make sense to encode the second constraint as one of minimizing some measure of the distance between the distributions given by the two learners.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"For irt the Good-Turing estimate just discussed gives us an estimate of p(unseen(f,) I f,)-the probability of observing a previously unseen instance of a construction in ft given that we know that we have a construction in f,.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"The confidence intervals are computed by bootstrap resampling for BLEU, and by standard significance testing for the manual scores, as described earlier in the paper.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"Making the ten judgements (2 types for 5 systems) takes on average 2 minutes.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Finally, as\u00c2\u00ad suming a simple bigram backoff model, we can derive the probability estimate for the particular unseen word i\u00c2\u00a51J1l.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"They return a value in the range [0,1], where 0 indicates neutrality and 1 indicates the strongest belief that the candidate and anaphor are coreferent.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"(2009) on Portuguese (Grac\u00c2\u00b8a et al.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"The extent to which this constraint is enforced varies greatly across existing methods.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"This approach needs a phrase as an initial seed and thus the possible relationships to be extracted are naturally limited.","label":0},{"source_text":"This paper talks about KenLM: Faster and Smaller Language Model Queries.","target_text":"Our TRIE implementation is designed to improve upon IRSTLM using a reverse trie with improved search, bit level packing, and stateful queries.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"Step 2.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"It is chosen such that the decisions it made in including or excluding constituents are most probable under the models for all of the parsers.","label":0},{"source_text":"NER is useful in many NLP applications such as information extraction, question answering, etc .Chieu and Ng have shown that the maximum entropy framework is able to use global information directly from various sources.","target_text":"As can be seen in Table 4, our training data is a lot less than those used by MENE and IdentiFinder3.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"Thus our proposed model is a proper model assigning probability mass to all (7r, L) pairs, where 7r is a parse tree and L is the one and only lattice that a sequence of characters (and spaces) W over our alpha-beth gives rise to.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"In the graphs, system scores are indicated by a point, the confidence intervals by shaded areas around the point.","label":0},{"source_text":"They showed the efficacy of graph-based label propagation for projecting part-of-speech information across languages.","target_text":"Central to our approach (see Algorithm 1) is a bilingual similarity graph built from a sentence-aligned parallel corpus.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"The standard measures for evaluating Penn Treebank parsing performance are precision and recall of the predicted constituents.","label":0},{"source_text":"Their method did not assume any knowledge about the target language, making it applicable to a wide array of resource-poor languages.","target_text":"Our full model (\u201cWith LP\u201d) outperforms the unsupervised baselines and the \u201cNo LP\u201d setting for all languages.","label":0},{"source_text":"From a DP-based solution to the traveling salesman problem, they present a novel technique to restrict the possible word reordering between source and target language in order to achieve an e\u00c3\u0086cient search algorithm.","target_text":"For a given partial hypothesis (C; j), the order in which the cities in C have been visited can be ignored (except j), only the score for the best path reaching j has to be stored.","label":0},{"source_text":"These clusters are computed using an SVD variant without relying on transitional structure.","target_text":"Statistics for all data sets are shown in Table 2.","label":0},{"source_text":"There is no global pruning.","target_text":"We are given a source string fJ 1 = f1:::fj :::fJ of length J, which is to be translated into a target string eI 1 = e1:::ei:::eI of length I. Among all possible target strings, we will choose the string with the highest probability: ^eI 1 = arg max eI 1 fPr(eI 1jfJ 1 )g = arg max eI 1 fPr(eI 1) Pr(fJ 1 jeI 1)g : (1) The argmax operation denotes the search problem, i.e. the generation of the output sentence in the target language.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"Our annotators pointed out that very often they made almost random decisions as to what relation to choose, and where to locate the boundary of a span.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"We can check, what the consequences of less manual annotation of results would have been: With half the number of manual judgements, we can distinguish about 40% of the systems, 10% less.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"We obtained positive results using a very simple phrase-based system in two different adaptation settings: using English\/French Europarl to improve a performance on a small, specialized medical domain; and using non-news portions of the NIST09 training material to improve performance on the news-related corpora.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"We further report SYNCS, the parsing metric of Cohen and Smith (2007), to facilitate the comparison.","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"2.2.4 Semantic Caseframe Expectations The third type of contextual role knowledge learned by BABAR is Semantic Caseframe Expectations.","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"In this case, this knowledge source reports that the candidate is not a viable antecedent for the anaphor.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"By considering derivation trees, and thus abstracting away from the details of the composition operation and the structures being manipulated, we are able to state the similarities and differences between the 'This work was partially supported by NSF grants MCS42-19116-CER, MCS82-07294 and DCR-84-10413, ARO grant DAA 29-84-9-0027, and DARPA grant N00014-85-K0018.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Participants were also provided with two sets of 2,000 sentences of parallel text to be used for system development and tuning.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Twentieth-century linguistic work on Chinese (Chao 1968; Li and Thompson 1981; Tang 1988,1989, inter alia) has revealed the incorrectness of this traditional view.","label":0},{"source_text":"The main result is that the combined system can recover non-projective dependencies with a precision sufficient to give a significant improvement in overall parsing accuracy, especially with respect to the exact match criterion, leading to the best reported performance for robust non-projective parsing of Czech.","target_text":"It is worth noting that, although nonprojective constructions are less frequent in DDT than in PDT, they seem to be more deeply nested, since only about 80% can be projectivized with a single lift, while almost 95% of the non-projective arcs in PDT only require a single lift.","label":0},{"source_text":"Two general approaches are presented and two combination techniques are described for each approach.","target_text":"The combining technique must act as a multi-position switch indicating which parser should be trusted for the particular sentence.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"Specifically, the lexicon is generated as: P (T , W |\u00cf\u0088) =P (T )P (W |T ) Word Type Features (FEATS): Past unsupervised POS work have derived benefits from features on word types, such as suffix and capitalization features (Hasan and Ng, 2009; Berg-Kirkpatrick et al.,2010).","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"Lexical and Morphological Ambiguity The rich morphological processes for deriving Hebrew stems give rise to a high degree of ambiguity for Hebrew space-delimited tokens.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"The annotator can then \u00e2\u0080\u009cclick away\u00e2\u0080\u009d those words that are here not used as connectives (such as the conjunction und (\u00e2\u0080\u0098and\u00e2\u0080\u0099) used in lists, or many adverbials that are ambiguous between connective and discourse particle).","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"In this paper, we offer broad insight into the underperformance of Arabic constituency parsing by analyzing the interplay of linguistic phenomena, annotation choices, and model design.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Chang of Tsinghua University, Taiwan, R.O.C., for kindly providing us with the name corpora.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"The algorithm works due to the fact that not all permutations of cities have to be considered explicitly.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"Pairwise comparison: We can use the same method to assess the statistical significance of one system outperforming another.","label":0},{"source_text":"The departure from the traditional token-based tagging approach allow them to explicitly capture type-level distributional properties of valid POS tag assignments as part of the model.","target_text":"While our method also enforces a singe tag per word constraint, it leverages the transition distribution encoded in an HMM, thereby benefiting from a richer representation of context.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"(2010) consistently outperforms ours on English, we obtain substantial gains across other languages.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"Given the limited number of judgements we received, we did not try to evaluate this.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"We received submissions from 14 groups from 11 institutions, as listed in Figure 2.","label":0},{"source_text":"The overall parsing accuracy obtained with the pseudo-projective approach is still lower than for the best projective parsers.","target_text":"Except for the left3The graphs satisfy all the well-formedness conditions given in section 2 except (possibly) connectedness.","label":0},{"source_text":"The experimental tests are carried out on the Verbmobil task which is a limited-domain spoken-language task.","target_text":"The resulting algorithm is depicted in Table 1.","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"There exist a few robust broad-coverage parsers that produce non-projective dependency structures, notably Tapanainen and J\u00a8arvinen (1997) and Wang and Harper (2004) for English, Foth et al. (2004) for German, and Holan (2004) for Czech.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"64 76.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"The following auxiliary quantity is defined: Qe0 (e; C; j) := probability of the best partial hypothesis (ei 1; bi 1), where C = fbkjk = 1; ; ig, bi = j, ei = e and ei\u00f4\u0080\u0080\u00801 = e0.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"Since the parameter and token components will remain fixed throughout experiments, we briefly describe each.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"Each round is composed of two stages; each stage updates one of the classifiers while keeping the other classifier fixed.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"2.","label":0},{"source_text":"There is no global pruning.","target_text":"The translation of one position in the source sentence may be postponed for up to L = 3 source positions, and the translation of up to two source positions may be anticipated for at most R = 10 source positions.","label":0},{"source_text":"This assumption, however, is not inherent to type-based tagging models.","target_text":"The corresponding token words w are drawn conditioned on t and \u00ce\u00b8.2 Our full generative model is given by: K P (\u00cf\u0086, \u00ce\u00b8|T , \u00ce\u00b1, \u00ce\u00b2) = n (P (\u00cf\u0086t|\u00ce\u00b1)P (\u00ce\u00b8t|T , \u00ce\u00b1)) t=1 The transition distribution \u00cf\u0086t for each tag t is drawn according to DIRICHLET(\u00ce\u00b1, K ), where \u00ce\u00b1 is the shared transition and emission distribution hyperparameter.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"While our method also enforces a singe tag per word constraint, it leverages the transition distribution encoded in an HMM, thereby benefiting from a richer representation of context.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Indeed there are several open issues.","label":0},{"source_text":"The fact that projective dependency parsers can never exactly reproduce the analyses found in non-projective tree banks is often neglected because of the relative scarcity of problematic constructions.","target_text":"As can be seen from the last column in Table 1, both Head and Head+Path may theoretically lead to a quadratic increase in the number of distinct arc labels (Head+Path being worse than Head only by a constant factor), while the increase is only linear in the case of Path.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"The baseline alignment model does not permit that a source word is aligned to two or more target words, e.g. for the translation direction from German toEnglish, the German compound noun 'Zahnarztter min' causes problems, because it must be translated by the two target words dentist's appointment.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"8 Volume 1, English language, 19961997, Format version 1, correction level 0 An ap ho r T e r r o r i s m R e c Pr F D i s a s t e r s R e c Pr F De f. NP s Pro no uns .43 .79 .55 .50 .72 .59 .42 .91 .58 .42 .82 .56 Tot al .46 .76 .57 .42 .87 .57 Table 2: General Knowledge Sources Table 4: Individual Performance of KSs for Terrorism Table 3: General + Contextual Role Knowledge Sources larger MUC4 and Reuters corpora.9 4.2 Experiments.","label":0},{"source_text":"Explanations for this phenomenon are relative informativeness of lexicalization, insensitivity to morphology and the effect of variable word order and these factors lead to syntactic disambiguation.","target_text":"Lattice parsing (Chappelier et al., 1999) is an alternative to a pipeline that prevents cascading errors by placing all segmentation options into the parse chart.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"Figure 3 shows a small fragment of the WFST encoding the dictionary, containing both entries forjust discussed, g:t\u00c2\u00a5 zhonglhua2 min2guo2 (China Republic) 'Republic of China,' and i\u00c2\u00a5inl.","label":0},{"source_text":"In this paper, the authors proposed an approach for instance-weighting phrase pairs in an out-of-domain corpus in order to improve in-domain performance.","target_text":"We describe a new approach to SMT adaptation that weights out-of-domain phrase pairs according to their relevance to the target domain, determined by both how similar to it they appear to be, and whether they belong to general language or not.","label":0},{"source_text":"This paper conducted research in the area of automatic paraphrase discovery.","target_text":"The second stage links sets which involve the same pairs of individual NEs.","label":0},{"source_text":"In this work, they take a more direct approach and treat a word type and its allowed POS tags as a primary element of the model.","target_text":"41.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"The evaluation framework for the shared task is similar to the one used in last year\u2019s shared task.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"In words, the judgements are normalized, so that the average normalized judgement per judge is 3.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"The number of judgements is additionally fragmented by our breakup of sentences into in-domain and out-of-domain.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"The table makes clear that enriching our grammar improves the syntactic performance as well as morphological disambiguation (segmentation and POS tagging) accuracy.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"Location list is processed into a list of unigrams and bigrams (e.g., New York).","label":0},{"source_text":"Bean and Riloff also used bootstrapping to extend their semantic compatibility model, proposed using caseframe network for anaphora resolution, information extraction patterns to identify contextual clues for determining compatibility between NPs.","target_text":"The goal of our research was to explore the use of contextual role knowledge for coreference resolution.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"2.2.1 The Caseframe Representation Information extraction (IE) systems use extraction patterns to identify noun phrases that play a specific role in 1 Our implementation only resolves NPs that occur in the same document, but in retrospect, one could probably resolve instances of the same existential NP in different documents too.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"Part-of-speech (POS) tag distributions are known to exhibit sparsity \u00e2\u0080\u0094 a word is likely to take a single predominant tag in a corpus.","label":0},{"source_text":"On the basis of that observation, they described a class of formalisms which they called Linear Context- Free Rewriting Systems (LCFRs), and showed they were recognizable in polynomial time and generated only semilinear languages.","target_text":"The following rule corresponds to the above derivation, where 71, , 7k are derived from the auxiliary trees , , fik, respectively. for all addresses n in some elementary tree at which 7' can be adjoined.","label":0},{"source_text":"Evaluation was done automatically using the BLEU score and manually on fluency and adequacy.","target_text":"For instance, in the recent IWSLT evaluation, first fluency annotations were solicited (while withholding the source sentence), and then adequacy annotations.","label":0},{"source_text":"The contextual rules are restricted and may not be applicable to every example, but the spelling rules are generally applicable and should have good coverage.","target_text":"Having found (spelling, context) pairs in the parsed data, a number of features are extracted.","label":0},{"source_text":"The texts were annotated with the RSTtool.","target_text":"Nonetheless there is no alternative to Reiche\u00e2\u0080\u0099s plan.","label":0},{"source_text":"The bias of automatic methods in favour of statistical systems seems to be less pronounced on out-of-domain test data.","target_text":"For the automatic evaluation, we used BLEU, since it is the most established metric in the field.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"This is done using a simple PCFG which is lexemebased.","label":0},{"source_text":"The authors use linguistic and annotation insights to develop a manually annotated grammar and evaluate it and finally provide a realistic evaluation in which segmentation is performed in a pipeline jointly with parsing.","target_text":"We weight edges using a unigram language model estimated with Good- Turing smoothing.","label":0},{"source_text":"Most IE researchers have been creating paraphrase knowledge by hand and specific tasks.","target_text":"Automatic paraphrase discovery is an important but challenging task.","label":0},{"source_text":"Vijay-Shankar et all considered the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate.","target_text":"Nodes are annotated by the name of the composition operation used at that step in the derivation.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"Not all the layers have been produced for all the texts yet.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"Any NLP application that presumes as input unrestricted text requires an initial phase of text analysis; such applications involve problems as diverse as machine translation, information retrieval, and text-to-speech synthesis (TIS).","label":0},{"source_text":"Their results show that their high performance NER use less training data than other systems.","target_text":"In this paper, we show that the maximum entropy framework is able to make use of global information directly, and achieves performance that is comparable to the best previous machine learning-based NERs on MUC6 and MUC7 test data.","label":0},{"source_text":"There are clustering approaches that assign a single POS tag to each word type.","target_text":"Once the lexicon has been drawn, the model proceeds similarly to the standard token-level HMM: Emission parameters \u00ce\u00b8 are generated conditioned on tag assignments T . We also draw transition parameters \u00cf\u0086.","label":0},{"source_text":"Human judges also pointed out difficulties with the evaluation of long sentences.","target_text":"Still, for about good number of sentences, we do have this direct comparison, which allows us to apply the sign test, as described in Section 2.2.","label":0},{"source_text":"It also incorporates the Good-Turing methodin estimating the likelihoods of previously unseen constructions, including morphological derivatives and personal names.","target_text":"In this paper we have argued that Chinese word segmentation can be modeled ef\u00c2\u00ad fectively using weighted finite-state transducers.","label":0},{"source_text":"It is probably the first analysis of Arabic parsing of this kind.","target_text":"Because these two words have identical complements, syntax rules are typically unhelpful for distinguishing between them.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"(2010)\u00e2\u0080\u0099s richest model: optimized via either EM or LBFGS, as their relative performance depends on the language.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"Although these existential NPs do not need a prior referent, they may occur multiple times in a document.","label":0},{"source_text":"It is well-known that English constituency parsing models do not generalize to other languages and treebanks.","target_text":"Third, we develop a human interpretable grammar that is competitive with a latent variable PCFG.","label":0},{"source_text":"The three parsers were trained and tuned by their creators on various sections of the WSJ portion of the Penn Treebank.","target_text":"None of the parsers produce parses with crossing brackets, so none of them votes for both of the assumed constituents.","label":0},{"source_text":"There is no global pruning.","target_text":"kann 7.nicht 8.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"We will refer to the semantic classes that co-occur with a caseframe as the semantic expectations of the caseframe.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"For each pair of judges consider one judge as the standard,.","label":0},{"source_text":"They proposed a single joint model for performing both morphological segmentation and syntactic disambiguation which bypasses the associated circularity.","target_text":"The Grammar Our parser looks for the most likely tree spanning a single path through the lattice of which the yield is a sequence of lexemes.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"4.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"Further, the probing hash table does only one random lookup per query, explaining why it is faster on large data.","label":0},{"source_text":"The resulting model is compact, efficiently learnable and linguistically expressive.","target_text":"Hyperparameter settings are sorted according to the median one-to-one metric over runs.","label":0},{"source_text":"The problem of coreference resolution has received considerable attention, including theoretical discourse models and supervised machine learning systems.","target_text":"2.2.2 The Caseframe Network The first type of contextual role knowledge that BABAR learns is the Caseframe Network (CFNet), which identifies caseframes that co-occur in anaphor\/antecedent resolutions.","label":0},{"source_text":"They extended previous work on discriminative weighting by using a finer granularity, focusing on the properties of instances rather than corpus components, and used simpler training procedure.","target_text":"In addition to using the simple features directly, we also trained an SVM classifier with these features to distinguish between IN and OUT phrase pairs.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"Figure 2 shows examples of lexical expectations that were learned for both domains.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"Smith estimates Lotus will make profit this quarter\u00e2\u0080\u00a6\u00e2\u0080\u009d.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"Xim, } associated with the ith example.","label":0},{"source_text":"Through parser combination we have reduced the precision error rate by 30% and the recall error rate by 6% compared to the best previously published result.","target_text":"IL+-1Proof: Assume a pair of crossing constituents appears in the output of the constituent voting technique using k parsers.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"As seen by the drop in average individual parser performance baseline, the introduced parser does not perform very well.","label":0},{"source_text":"They focused on phrases which two Named Entities, and proceed in two stages.","target_text":"We have checked if there are similar verbs in other major domains, but this was the only one.","label":0},{"source_text":"They plan on extending instance-weighting to other standard SMT components and capture the degree of generality of phrase pairs.","target_text":"A final alternate approach would be to combine weighted joint frequencies rather than conditional estimates, ie: cI(s, t) + w,\\(s, t)co(, s, t), suitably normalized.5 Such an approach could be simulated by a MAP-style combination in which separate 0(t) values were maintained for each t. This would make the model more powerful, but at the cost of having to learn to downweight OUT separately for each t, which we suspect would require more training data for reliable performance.","label":0},{"source_text":"They used graph-based label propagation for cross-lingual knowledge transfer and used the projected labels as features in an unsupervised model.","target_text":"It would have therefore also been possible to use the integer programming (IP) based approach of Ravi and Knight (2009) instead of the feature-HMM for POS induction on the foreign side.","label":0},{"source_text":"Using a treebank grammar, the data-driven lexicon, and a linguistically motivated unknown-tokens handling technique their model outperformed previous pipelined, integrated or factorized systems for Hebrew morphological and syntactic processing, yielding an error reduction of 12% over the best published results so far.","target_text":"Current state-of-the-art broad-coverage parsers assume a direct correspondence between the lexical items ingrained in the proposed syntactic analyses (the yields of syntactic parse-trees) and the spacedelimited tokens (henceforth, \u2018tokens\u2019) that constitute the unanalyzed surface forms (utterances).","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"The sentence-selection approach is crude in that it imposes a binary distinction between useful and non-useful parts of OUT.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"(Yarowsky 95) describes an algorithm for word-sense disambiguation that exploits redundancy in contextual features, and gives impressive performance.","label":0},{"source_text":"Koehn and Monz carried out an extensive manual and automatic evaluation of machine translation performance on European language pairs.","target_text":"Unfortunately, we have much less data to work with than with the automatic scores.","label":0},{"source_text":"Replacing this with a ranked evaluation seems to be more suitable.","target_text":"It rewards matches of n-gram sequences, but measures only at most indirectly overall grammatical coherence.","label":0},{"source_text":"they evaluate various specific aspects of the segmentation, as well as the overall segmentation performance.","target_text":"The segmentation chosen is the best path through the WFST, shown in (d).","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"This causes a problem for reverse trie implementations, including SRILM itself, because it leaves n+1-grams without an n-gram node pointing to them.","label":0},{"source_text":"While many systems had similar performance, the results offered interesting insights, especially, about the relative performance of statistical and rule-based systems.","target_text":"This is less than the 694 judgements 2004 DARPA\/NIST evaluation, or the 532 judgements in the 2005 DARPA\/NIST evaluation.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"We show that noun-noun vs. discourse-level coordination ambiguity in Arabic is a significant source of parsing errors (Table 8c).","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"On one end of the spectrum are clustering approaches that assign a single POS tag to each word type (Schutze, 1995; Lamar et al., 2010).","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"This is not completely surprising, since all systems use very similar technology.","label":0},{"source_text":"This architecture provides a uniform framework in which it is easy to incorporate not only listed dictionary entries but also morphological derivatives, and models for personal names and foreign names in transliteration.","target_text":"42 nator, the N31s can be measured well by counting, and we replace the expectation by the observation.","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"The NP is a complement to a preposition, which is the head of a PP.","label":0},{"source_text":"the \"Potsdam Commentary Corpus\" or PCC consists of 170 commentaries from Ma\u00a8rkische Allgemeine Zeitung, a German regional daily.","target_text":"The state in future has not enough work for its many teachers.","label":0},{"source_text":"The PROBING data structure uses linear probing hash tables and is designed for speed.","target_text":"However, this optimistic search would not visit the entries necessary to store backoff information in the outgoing state.","label":0},{"source_text":"The approach assumes that the word reordering is restricted to a few positions in the source sentence.","target_text":"This measure has the advantage of being completely automatic.","label":0},{"source_text":"They employed a PCFG-based generative framework to make both syntactic and morphological disambiguation decisions which were not only theoretically clean and linguistically justified but also probabilistically appropriate and empirically sound.","target_text":"The form fmnh, for example, can be understood as the verb \u201clubricated\u201d, the possessed noun \u201cher oil\u201d, the adjective \u201cfat\u201d or the verb \u201cgot fat\u201d.","label":0},{"source_text":"They believe that global context is useful in most languages, as it is a natural tendency for authors to use abbreviations on entities already mentioned previously.","target_text":"The overall performance of the LTG system was outstanding, but the system consists of a sequence of many hand-coded rules and machine-learning modules.","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"Here, annotation proceeds in two phases: first, the domains and the units of IS are marked as such.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"This representation gives ir, an appropriate morphological decomposition, pre\u00c2\u00ad serving information that would be lost by simply listing ir, as an unanalyzed form.","label":0},{"source_text":"They incorporated instance-weighting into a mixture-model framework, and found that it yielded consistent improvements over a wide range of baselines.","target_text":"Log-linear combination (loglin) improves on this in all cases, and also beats the pure IN system.","label":0},{"source_text":"Other kinds of productive word classes, such as company names, abbreviations,and place names can easily be handled given appropriate models.","target_text":"In a few cases, the criteria for correctness are made more explicit.","label":0},{"source_text":"Their empirical results demonstrate that the type-based tagger rivals state-of-the-art tag-level taggers which employ more sophisticated learning mechanisms to exploit similar constraints.","target_text":"Feature-based HMM Model (Berg- Kirkpatrick et al., 2010): The KM model uses a variety of orthographic features and employs the EM or LBFGS optimization algorithm; Posterior regulariation model (Grac\u00c2\u00b8a et al., 2009): The G10 model uses the posterior regular- ization approach to ensure tag sparsity constraint.","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"Moses keeps language models and many other resources in static variables, so these are still resident in memory.","label":0},{"source_text":"Due to many similarly performing systems, the author was not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"One annotator suggested that this was the case for as much as 10% of our test sentences.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"BerkeleyLM uses states to optimistically search for longer n-gram matches first and must perform twice as many random accesses to retrieve backoff information.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"For a sequence of hanzi that is a possible name, we wish to assign a probability to that sequence qua name.","label":0},{"source_text":"For the perplexity and translation tasks, we used SRILM to build a 5-gram English language model on 834 million tokens from Europarl v6 (Koehn, 2005) and the 2011 Workshop on Machine Translation News Crawl corpus with duplicate lines removed.","target_text":"For RandLM and IRSTLM, the effect of caching can be seen on speed and memory usage.","label":0},{"source_text":"It is annotated with several data: morphology, syntax, rhetorical structure, connectors, correference and informative structure.","target_text":"In (Reitter, Stede 2003) we went a different way and suggested URML5, an XML format for underspecifying rhetorical structure: a number of relations can be assigned instead of a single one, competing analyses can be represented with shared forests.","label":0},{"source_text":"The model incorporates various recent techniques for incorporating and manipulating linguistic knowledge using finite-state transducers.","target_text":"Clearly, for judges h and h taking h as standard and computing the precision and recall for Jz yields the same results as taking h as the standard, and computing for h, 14 All evaluation materials, with the exception of those used for evaluating personal names were drawn.","label":0},{"source_text":"The corpus was annoted with different linguitic information.","target_text":"We follow the guidelines developed in the TIGER project (Brants et al. 2002) for syntactic annotation of German newspaper text, using the Annotate3 tool for interactive construction of tree structures.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"The first is an evaluation of the system's ability to mimic humans at the task of segmenting text into word-sized units; the second evaluates the proper-name identification; the third measures the performance on morphological analysis.","label":0},{"source_text":"The authors show that PATB is similar to other tree-banks but that annotation consistency remains low.","target_text":"markBaseNP indicates these non-recursive nominal phrases.","label":0},{"source_text":"The overall performance of their joint framework demonstrated that a probability distribution obtained over mere syntactic contexts using a Treebank grammar and a data-driven lexicon outperformed upper bounds proposed by previous joint disambiguation systems and achieved segmentation and parsing results on a par with state-of-the-art standalone applications results.","target_text":"Our smoothing procedure takes into account all the aforementioned aspects and works as follows.","label":0},{"source_text":"Finally, several coreference systems have successfully incorporated anaphoricity determination modules.","target_text":"The contextual role knowledge that BABAR uses for coreference resolution is derived from this caseframe data.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"In order to pass these constraints onto the parser, the lexical rules in the grammar are of the form pi \u2014* (si, pi) Parameter Estimation The grammar probabilities are estimated from the corpus using simple relative frequency estimates.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"In many cases, there is an even stronger restriction: over large portions of the source string, the alignment is monotone.","label":0},{"source_text":"This topic has been getting more attention, driven by the needs of various NLP applications.","target_text":"shows some keywords with their scores.","label":0},{"source_text":"The authors cluster NE instance pairs based on the words in the context using bag-of-words methods.","target_text":"For example, in the CC-domain, 96 keywords are found which have TF\/ITF scores above a threshold; some of them are shown in Figure 3.","label":0},{"source_text":"Their work is closely related to recent approaches that incorporate the sparsity constraint into the POS induction process.","target_text":"Table 5 provides insight into the behavior of different models in terms of the tagging lexicon they generate.","label":0},{"source_text":"They have made use of local and global features to deal with the instances of same token in a document.","target_text":"If is not initCaps, then (not-initCaps, ) is set to 1.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"2.","label":0},{"source_text":"Across eight European languages, their approach results in an average absolute improvement of 10.4% over a state-of-the-art baseline, and 16.7% over vanilla hidden Markov models induced with the Expectation Maximization algorithm.","target_text":"For each language under consideration, Petrov et al. (2011) provide a mapping A from the fine-grained language specific POS tags in the foreign treebank to the universal POS tags.","label":0},{"source_text":"In this paper the authors present a stochastic finite-state model for segmenting Chinese text into words.","target_text":"All notions of word, with the exception of the orthographic word, are as relevant in Chinese as they are in English, and just as is the case in other languages, a word in Chinese may correspond to one or more symbols in the orthog 1 For a related approach to the problem of word-segrnention in Japanese, see Nagata (1994), inter alia..","label":0},{"source_text":"All the texts were annotated by two people.","target_text":"Consequently, we implemented our own annotation tool ConAno in Java (Stede, Heintze 2004), which provides specifically the functionality needed for our purpose.","label":0},{"source_text":"This paper presents a maximum entropy-based named entity recognizer (NER).","target_text":"He used an additional maximum entropy classifier that tries to correct mistakes by using reference resolution.","label":0},{"source_text":"However, using the top-level semantic classes of WordNet proved to be problematic as the class distinctions are too coarse.","target_text":"Sometimes, however, these beliefs can be contradictory.","label":0},{"source_text":"Due to many similarly performing systems, they are not able to draw strong conclusions on the question of correlation of manual and automatic evaluation metrics.","target_text":"Since the inclusion of out-ofdomain test data was a very late decision, the participants were not informed of this.","label":0},{"source_text":"The authors in this paper describe a search procedure for statistical machine translation (MT) based on dynamic programming (DP).","target_text":"Source sentence words are aligned with hypothesized target sentence words, where the choice of a new source word, which has not been aligned with a target word yet, is restricted1.","label":0},{"source_text":"A beam search concept is applied as in speech recognition.","target_text":"However, dynamic programming can be used to find the shortest tour in exponential time, namely in O(n22n), using the algorithm by Held and Karp.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"The morphological anal\u00c2\u00adysis itself can be handled using well-known techniques from finite-state morphol 9 The initial estimates are derived from the frequencies in the corpus of the strings of hanzi making up.","label":0},{"source_text":"The approach has been successfully tested on the 8 000-word Verbmobil task.","target_text":"Additionally, it works about 3 times as fast as the IBM style search.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1&#8211;5 seemed to be very hard to perform.","target_text":"The bias of automatic methods in favor of statistical systems seems to be less pronounced on out-of-domain test data.","label":0},{"source_text":"They showed better grammars to improve performance on both morphological and syntactic tasks, providing support for the advantage of a joint framework over pipelined or factorized ones.","target_text":"The input for the segmentation task is however highly ambiguous for Semitic languages, and surface forms (tokens) may admit multiple possible analyses as in (BarHaim et al., 2007; Adler and Elhadad, 2006).","label":0},{"source_text":"The AdaBoost algorithm was developed for supervised learning.","target_text":"Unfortunately, Yarowsky's method is not well understood from a theoretical viewpoint: we would like to formalize the notion of redundancy in unlabeled data, and set up the learning task as optimization of some appropriate objective function.","label":0},{"source_text":"The first method builds on results from (Yarowsky 95) and (Blum and Mitchell 98).","target_text":"Yarowsky-cautious does not separate the spelling and contextual features, but does have a limit on the number of rules added at each stage.","label":0},{"source_text":"they showed that the average agreement among the human judges is .76, and the average agreement between ST(system) and the humans is .75, or about 99% of the interhuman agreement.","target_text":"However, it is almost universally the case that no clear definition of what constitutes a \"correct\" segmentation is given, so these performance measures are hard to evaluate.","label":0},{"source_text":"They found replacing it with a ranked evaluation to be more suitable.","target_text":"However, it is also mostly political content (even if not focused on the internal workings of the European Union) and opinion.","label":0},{"source_text":"In this paper, Ben and Riloff present a coreference resolver called BABAR that focuses on the use of contextual-role knowledge for coreference resolution.","target_text":"In recent years, coreference resolvers have been evaluated as part of MUC6 and MUC7 (MUC7 Proceedings, 1998).","label":0},{"source_text":"BABAR's performance in both domains of terrorism and natural disaster, and the contextual-role knowledge in pronouns have shown successful results.","target_text":"However, using the top-level semantic classes of WordNet proved to be problematic because the class distinctions are too coarse.","label":0},{"source_text":"Combining multiple highly-accurate independent parsers yields promising results.","target_text":"Exploiting Diversity in Natural Language Processing: Combining Parsers","label":0},{"source_text":"Throughout this paper we compare with several packages: SRILM 1.5.12 (Stolcke, 2002) is a popular toolkit based on tries used in several decoders.","target_text":"These packages are further described in Section 3.","label":0},{"source_text":"This paper discusses the Potsdam Commentary Corpus, a corpus of german assembeled by potsdam university.","target_text":"Thus it is possible, for illustration, to look for a noun phrase (syntax tier) marked as topic (information structure tier) that is in a bridging relation (co-reference tier) to some other noun phrase.","label":0},{"source_text":"The manual evaluation of scoring translation on a graded scale from 1\u00e2\u0080\u00935 seems to be very hard to perform.","target_text":"It was also proposed to allow annotators to skip sentences that they are unable to judge.","label":0},{"source_text":"This paper presents methods to query N-gram language models, minimizing time and space costs.","target_text":"This differs from other implementations (Stolcke, 2002; Pauls and Klein, 2011) that use hash tables as nodes in a trie, as explained in the next section.","label":0},{"source_text":"The features were weighted within a logistic model that gave an overall weight that was applied to the phrase pair and MAP-smoothed relative-frequency estimates which were combined linearly with relative-frequency estimates from an in-domain phrase table.","target_text":"The iw all map variant uses a non-0 y weight on a uniform prior in p,,(s t), and outperforms a version with y = 0 (iw all) and the \u201cflattened\u201d variant described in section 3.2.","label":0}]