sentence1
stringlengths 16
446
| sentence2
stringlengths 14
436
|
|---|---|
morphological disambiguation is a useful first step for higher level analysis of any language but it is especially critical for agglutinative languages like turkish , czech , hungarian , and finnish . | morphological disambiguation is a well studied problem in the literature , but lstm-based contributions are still relatively scarce . |
we evaluate the performance of different translation models using both bleu and ter metrics . | we optimise the feature weights of the model with minimum error rate training against the bleu evaluation metric . |
dp beam search for phrase-based smt was described by koehn et al , extending earlier work on word-based smt . | belz and kow proposed another smt based nlg system which made use of the phrase-based smt model . |
we investigate the need for bigram alignment models and the benefit of supervised alignment techniques . | we have investigated the need for bigram alignment models and the benefit of supervised alignment techniques in g2p . |
hamilton et al report almost perfect accuracy for the procrustes transformation when detecting the direction of semantic change . | hamilton et al measured the variation between models by observing semantic change using diachronic corpora . |
word sense disambiguation ( wsd ) is the task of determining the meaning of a word in a given context . | word sense disambiguation ( wsd ) is a fundamental task and long-standing challenge in natural language processing ( nlp ) . |
for our experiments , we use 40,000 sentences from europarl for each language pair following the basic setup of tiedemann . | we perform the analysis with data from 110 different language pairs drawn from the europarl project . |
given a set of question-answer pairs as the development set , we use the minimum error rate training algorithm to tune the feature weights δ½ m i in our proposed model . | to optimize the feature weights for our model , we use viterbi envelope semiring training , which is an implementation of the minimum error rate training algorithm for training with an arbitrary loss function . |
garg and henderson proposed a stack long short-term memory approach to supervised dependency parsing . | garg and henderson used rbm in a similar approach to dependency parsing . |
sentences are tagged and parsed using the stanford dependency parser . | dependency parses are obtained from the stanford parser . |
princeton wordnet 1 is an english lexical database that groups nouns , verbs , adjectives and adverbs into sets of cognitive synonyms , which are named as synsets . | princeton wordnet is an english lexical database that groups nouns , verbs , adjectives and adverbs into sets of cognitive synonyms , which are named as synsets . |
we define the position set of math-w-7-11-0-40 , denoted by math-w-7-11-0-44 , as the set of all positions math-w-7-11-0-53 . | given an alphabet math-w-2-6-2-60 , we write math-w-2-6-2-64 for the set of all ( finite ) strings over math-w-2-6-2-76 . |
the standard approach to word alignment from sentence-aligned bitexts has been to construct models which generate sentences of one language from the other , then fitting those generative models with em . | the standard approach to word alignment is to construct directional generative models , which produce a sentence in one language given the sentence in another language . |
we used the moses toolkit to build mt systems using various alignments . | we used the moses toolkit for performing statistical machine translation . |
documents show that our model is effective in exploiting both source and target document context , and statistically significantly outperforms the previous work in terms of bleu and meteor . | the experimental results and analysis demonstrate that our model is effective in exploiting both source and target document context , and statistically significantly outperforms the previous work in terms of bleu and meteor . |
dredze et al , show that domain adaptation is hard for dependency parsing based on results in the conll 2007 shared task . | dredze et al showed that many of the parsing errors in domain adaptation tasks may come from inconsistencies between the annotations of training resources . |
neural machine translation has become the primary paradigm in machine translation literature . | neural machine translation has recently gained popularity in solving the machine translation problem . |
favorable compares with a tomita parser and a chart parser parsing time when run on the same grammar and lexicon . | the parsing time favorable compares with a tomita parser and a chart parser parsing time when run on the same grammar and lexicon . |
word sense disambiguation ( wsd ) is the problem of assigning a sense to an ambiguous word , using its context . | word sense disambiguation ( wsd ) is the task of determining the correct meaning or sense of a word in context . |
in this paper we describe the system submitted for the semeval 2014 sentiment analysis in twitter task ( task 9 subtask b ) . | in this paper we describe the system submitted for the semeval 2014 task 9 ( sentiment analysis in twitter ) subtask b . |
in an example shown above , β sad β is an emotion word , and the cause of β sad β is β . | in an example shown above , β sad β is an emotion word , and the cause of β sad β is β i lost my phone β . |
both the structure and semantic constraints from knowledge bases can be easily exploited during parsing . | it has been shown that structure and semantic constraints are effective for enhancing semantic parsing . |
semantic role labeling ( srl ) is a major nlp task , providing a shallow sentence-level semantic analysis . | semantic role labeling ( srl ) is the task of identifying the predicate-argument structure of a sentence . |
drezde et al applied structural correspondence learning to the task of domain adaptation for sentiment classification of product reviews . | blitzer et al investigate domain adaptation for sentiment classifiers , focusing on online reviews for different types of products . |
for training we use the adam optimizer with default values and mini-batches of 10 examples . | we train the model using the adam optimizer with the default hyper parameters . |
for other researchers who wish to use our indexing machinery , it has been made available as free software . | our own implementation will be made available to other researchers as open source . |
the phrase-based translation model has demonstrated superior performance and been widely used in current smt systems , and we employ our implementation on this translation model . | our work can be applied to any statistical machine translation paradigm and we will present results on a standard phrase-based translation system and a hierarchical phrase-based translation system . |
shen et al , 2008 ) presents a string-to-dependency model , which restricts the target side of each hierarchical rule to be a well-formed dependency tree fragment , and employs a dependency language model to make the output more grammatically . | shen et al , 2008 ) extends the hierarchical phrase-based model and present a string-to-dependency model , which employs string-to-dependency rules whose source side are string and the target as well-formed dependency structures . |
the grammatical framework for the krg is head-driven phrase structure grammar , a non-derivational , constraintbased , and surface-oriented grammatical architecture . | the lingo grammar matrix is situated theoretically within head-driven phrase structure grammar , a lexicalist , constraint-based framework . |
we then use the stanford sentiment classifier developed by socher et al to automatically assign sentiment labels to translated tweets . | we then use the stanford sentiment classifier to automatically assign sentiment labels to translated tweets . |
the scores of participants are in table 10 in terms of bleu and f 1 scores . | performance is measured based on the bleu scores , which are reported in table 4 . |
we use the pre-trained glove 50-dimensional word embeddings to represent words found in the glove dataset . | the word vectors of vocabulary words are trained from a large corpus using the glove toolkit . |
coreference resolution is the task of clustering a sequence of textual entity mentions into a set of maximal non-overlapping clusters , such that mentions in a cluster refer to the same discourse entity . | coreference resolution is the task of automatically grouping references to the same real-world entity in a document into a set . |
a 5-gram language model was built using srilm on the target side of the corresponding training corpus . | further , we apply a 4-gram language model trained with the srilm toolkit on the target side of the training corpus . |
relation extraction is a fundamental task in information extraction . | relation extraction ( re ) is the process of generating structured relation knowledge from unstructured natural language texts . |
we propose a novel framework for speech disfluency detection based on integer linear programming ( ilp ) . | we present a novel two-stage technique for detecting speech disfluencies based on integer linear programming ( ilp ) . |
to rerank the candidate texts , we used a 5-gram language model trained on the europarl corpus using kenlm . | after standard preprocessing of the data , we train a 3-gram language model using kenlm . |
coreference resolution is the problem of identifying which mentions ( i.e. , noun phrases ) refer to which real-world entities . | coreference resolution is the task of clustering referring expressions in a text so that each resulting cluster represents an entity . |
summarization is a classic text processing problem . | summarization is the process of condensing a source text into a shorter version while preserving its information content . |
ambiguity is a problem in any natural language processing system . | ambiguity is the task of building up multiple alternative linguistic structures for a single input ( cite-p-13-1-8 ) . |
using statistics from both standard and learner corpora , it generates plausible distractors . | focusing on prepositions , the system generates distractors based on error statistics compiled from learner corpora . |
we review prior work on topic modeling for document collections and studies of social media like political blogs . | in this paper we applied several probabilistic topic models to discourse within political blogs . |
we use the logistic regression classifier in the skll package , which is based on scikit-learn , optimizing for f 1 score . | for the feature-based system we used logistic regression classifier from the scikit-learn library . |
zou et al learn bilingual word embeddings by designing an objective function that combines unsupervised training with bilingual constraints based on word alignments . | the bilingual embedding research origins in the word embedding learning , upon which zou et al utilize word alignments to constrain translational equivalence . |
we used glove vectors trained on common crawl 840b 4 with 300 dimensions as fixed word embeddings . | we used the 300-dimensional glove word embeddings learned from 840 billion tokens in the web crawl data , as general word embeddings . |
in this paper , we propose a method for slu based on generative and discriminative models . | in this paper , we propose discriminative reranking of concept annotation to jointly exploit generative and discriminative models . |
we used 5-gram models , estimated using the sri language modeling toolkit with modified kneser-ney smoothing . | we also use a 4-gram language model trained using srilm with kneser-ney smoothing . |
our baseline is an in-house phrase-based statistical machine translation system very similar to moses . | our translation system is an in-house phrasebased system analogous to moses . |
bleu has long been shown not to correlate well with human judgment on translation quality . | bleu exhibits a high correlation with human judgments of translation quality when measuring on large sections of text . |
relation extraction is a subtask of information extraction that finds various predefined semantic relations , such as location , affiliation , rival , etc. , between pairs of entities in text . | relation extraction is the task of detecting and characterizing semantic relations between entities from free text . |
we conducted baseline experiments for phrasebased machine translation using the moses toolkit . | we used the moses toolkit for performing statistical machine translation . |
conjecture and empirically show that entailment graphs exhibit a β tree-like β property , i . e . , that they can be reduced into a structure similar to a directed forest . | we first identify that entailment graphs exhibit a β tree-like β property and are very similar to a novel type of graph termed forest-reducible graph . |
we used 300-dimensional pre-trained glove word embeddings . | we use pre-trained vectors from glove for word-level embeddings . |
automatic evaluation metrics , such as the bleu score , were crucial ingredients for the advances of machine translation technology in the last decade . | during the last decade , automatic evaluation metrics have helped researchers accelerate the pace at which they improve machine translation systems . |
twitter is a huge microblogging service with more than 500 million tweets per day from different locations of the world and in different languages ( cite-p-8-1-9 ) . | twitter is a widely used microblogging platform , where users post and interact with messages , β tweets β . |
we began our study by consulting the 51,558 parsed sentences of the wsj corpus . | we used sections 0 to 12 of the wsj part of the penn treebank with a total of 24,618 sentences for our experiments . |
narayanan et al proposed a method for sentiment classification targeting conditional sentences . | narayanan et al discuss a pos-based approach for identifying conditional types for the task of sentiment analysis . |
the nnlm weights are optimized as the other feature weights using minimum error rate training . | the feature weights are tuned with minimum error-rate training to optimise the character error rate of the output . |
vectorial representations derived from large current events datasets such as google news have been shown to perform well on word similarity tasks . | vectorial representations of words derived from large current events datasets have been shown to perform well on word similarity tasks . |
we used the phrase-based smt in moses 5 for the translation experiments . | we used the phrasebased translation system in moses 5 as a baseline smt system . |
amr is a formalism of sentence semantic structure by directed , acyclic , and rooted graphs , in which semantic relations such as predicate-argument relations and noun-noun relations are expressed . | an amr is a graph with nodes representing the concepts of the sentence and edges representing the semantic relations between them . |
we use the sentiment pipeline of stanford corenlp to obtain this feature . | for part of speech tagging and dependency parsing of the text , we used the toolset from stanford corenlp . |
in this paper , i have demonstrated how to build an entailment system from mrs graph alignment , combined with heuristic β robust β . | in this paper , i examine the benefits and possible disadvantages of using rich semantic representations as the basis for entailment recognition . |
that initialized em , improves parsing accuracy from 90 . 2 % to 91 . 8 % on english , and from 80 . 3 % to 84 . 5 % on german . | despite its simplicity , a product of eight automatically learned grammars improves parsing accuracy from 90.2 % to 91.8 % on english , and from 80.3 % to 84.5 % on german . |
gulordava and baroni consider the identification of diachronic changes in meaning from an n-gram database , but in contrast to sagi et al and cook and stevenson , do not focus on specific types of semantic change . | gulordava and baroni identify diachronic sense change in an n-gram database , but using a model that is not restricted to any particular type of semantic change . |
lmbr decoding can also be used as an effective framework for multiple lattice combination . | linearised lattice minimum bayes-risk decoding can also be used as an effective framework for multiple lattice combination . |
the availability of a large typology database makes it possible to take computational approaches to this area of study . | fortunately , the publication of a large typology database made it possible to take computational approaches to this area of study . |
with this method , the correlation rate reached 0 . 7667 , which represent the best score among the different submitted methods involved in the arabic monolingual sts task . | lim-lig system achieves a pearsons correlation of 0.74633 , ranking 2nd among all participants in the arabic monolingual pairs sts task organized within the semeval 2017 evaluation campaign . |
to this end , we design novel features based on citation network information and use them in conjunction with traditional features for keyphrase extraction . | to this end , we design novel features for keyphrase extraction based on citation context information and use them in conjunction with traditional features in a supervised probabilistic framework . |
we present a novel model of transliteration mining . | we presented a novel model to automatically mine transliteration pairs . |
in this paper , we propose using a constrained word lattice , which encodes input phrases and tm constraints . | in this paper , we propose a constrained word lattice to combine smt and tm at phrase-level . |
twitter is a microblogging site where people express themselves and react to content in real-time . | twitter is a huge microblogging service with more than 500 million tweets per day from different locations of the world and in different languages ( cite-p-10-1-6 ) . |
active learning is a machine learning approach to achieving high-accuracy with a small amount of labels by letting the learning algorithm choose instances to be labeled . | active learning is a promising way for sentiment classification to reduce the annotation cost . |
in this shared task , we employ the word embeddings model to reflect paradigmatic relationships between words . | we employ a neural method , specifically the continuous bag-of-words model to learn high-quality vector representations for words . |
it is much more efficient than the viterbi algorithm when dealing with a large number of labels . | viterbi decoding is , however , prohibitively slow when the label set is large , because its time complexity is quadratic in the number of labels . |
chung and gildea reported that the automatic insertion of empty categories improved the accuracy of phrased-based machine translation . | chung and gildea reported their recover of empty categories improved the accuracy of machine translation both in korean and in chinese . |
we use the logistic regression implementation of liblinear wrapped by the scikit-learn library . | we used the support vector machine implementation from the liblinear library on the test sets and report the results in table 4 . |
we used the bleu score to evaluate the translation accuracy with and without the normalization . | we computed the translation accuracies using two metrics , bleu score , and lexical accuracy on a test set of 30 sentences . |
for decoding , we used moses with the default options . | for the phrase based system , we use moses with its default settings . |
in previous work , hatzivassiloglou and mckeown propose a method to identify the polarity of adjectives . | hatzivassiloglou and mckeown proposed a method for identifying word polarity of adjectives . |
in this paper , we propose a novel framework , companion teaching , to include a human teacher in the dialogue policy training loop . | we propose a novel framework , companion teaching , to include a human teacher in the online dialogue policy training loop to address the cold start problem . |
however , aspect extraction is a complex task that also requires fine-grained domain embeddings . | aspect extraction is a task to abstract the common properties of objects from corpora discussing them , such as reviews of products . |
rozovskaya and roth further demonstrate that the models perform better when they use knowledge about error patterns of the non-native writers . | finally , rozovskaya and roth found that a classifier outperformed a language modeling approach on different data , making it unclear which approach is best . |
experiments show that our model achieves state-of-the-art f-score . | results show that our model outperforms previous state-of-the-art systems . |
question answering ( qa ) is a long-standing challenge in nlp , and the community has introduced several paradigms and datasets for the task over the past few years . | question answering ( qa ) is a challenging task that draws upon many aspects of nlp . |
as a model learning method , we adopt the maximum entropy model learning method . | we use the mallet implementation of a maximum entropy classifier to construct our models . |
among them , twitter is the most popular service by far due to its ease for real-time sharing of information . | twitter consists of a massive number of posts on a wide range of subjects , making it very interesting to extract information and sentiments from them . |
semantic role labeling ( srl ) is a task of analyzing predicate-argument structures in texts . | semantic role labeling ( srl ) is a form of shallow semantic parsing whose goal is to discover the predicate-argument structure of each predicate in a given input sentence . |
and consequently key phrases tend to have close semantics . | as a matter of fact , key phrases often have close semantics to title phrases . |
the syntax-based statistical machine translation models use rules with hierarchical structures as translation knowledge , which can capture long-distance reorderings . | to solve this , syntax-based models take tree structures into consideration to learn translation patterns by using non-terminals for generalization . |
in argument reconstruction , the induced roles largely correspond to roles defined in annotated resources . | when estimated jointly on unlabeled data , roles induced by the model mostly corresponds to roles defined in existing resources by annotators . |
by integrating the two components into an existing amr parser , our parser is able to outperform state-of-the-art amr parsers . | we show integrating the two components into an existing amr parser results in consistently better performance over the state of the art on various datasets . |
the language model is trained and applied with the srilm toolkit . | this means in practice that the language model was trained using the srilm toolkit . |
it has been empirically shown that word embeddings could capture semantic and syntactic similarities between words . | these embeddings provide a nuanced representation of words that can capture various syntactic and semantic properties of natural language . |
in phrase-based smt models , phrases are used as atomic units for translation . | in phrase-based smt , words may be grouped together to form so-called phrases . |
phrase-based statistical machine translation models have achieved significant improvements in translation accuracy over the original ibm word-based model . | phrase-based statistical translation systems are currently providing excellent results in real machine translation tasks . |
in a low-resource setting , we design a multitask learning approach that utilizes parallel data of a third language , called the pivot language . | we present a multi-task learning approach that jointly trains three word alignment models over disjoint bitexts of three languages : source , target and pivot . |
this paper describes limsi β s submission to the conll 2017 ud shared task , which is focused on small treebanks , and how to improve low-resourced parsing . | this paper describes limsi β s submission to the conll 2017 ud shared task ( cite-p-20-3-5 ) , dedicated to parsing universal dependencies ( cite-p-20-1-10 ) on a wide array of languages . |
we estimated unfiltered 5-gram language models using lmplz and loaded them with kenlm . | for all systems , we trained a 6-gram language model smoothed with modified kneser-ney smoothing using kenlm . |
we follow a previous attempt to use a sequence-to-sequence learning model augmented with the attention mechanism . | following previous work , we believe that using sequential information rather than a bag-of-words model would help improve performance . |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.