HHousen/ParaSCI · Datasets at Hugging Face

sentence1 stringlengths 16 446	sentence2 stringlengths 14 436
in this paper , we examine user adaptation to the system ’ s lexical and syntactic choices in the context of the deployed .	we confirm prior results showing that users adapt to the system ’ s lexical and syntactic choices .
adwords features can be used for defining message persuasiveness .	adwords gives us an appropriate context for evaluating persuasive messages .
for faster training , we employ an efficient parallel training strategy proposed by mcdonald et al .	we make use of a distributed training strategy for the structured perceptron that was first introduced in mcdonald et al .
toutanova and moore addressed the phonetic substitution problem by extending the initial letter-to-phone model .	toutanova and moore improved this approach by extending the error model with phonetic similarities over words .
but the neural networks leverage these features for improving tagging .	this observation is evidence that the neural network can find good representations for pos tagging .
in general , the use of modifier structures and the associated semantic interpretation component permits a good treatment of scoping problems involving coordination .	the system 's semantic interpretation component can in particular deal with scoping problems involving coordination .
relation extraction is the task of predicting attributes and relations for entities in a sentence ( zelenko et al. , 2003 ; bunescu and mooney , 2005 ; guodong et al. , 2005 ) .	relation extraction ( re ) is the task of extracting semantic relationships between entities in text .
we describe an endto-end generation model that performs content selection and surface realization .	similar to cite-p-11-1-0 , we also present an endto-end system that performs content selection and surface realization .
sentiment detection and classification has received considerable attention .	sentiment classification has seen a great deal of attention .
the 位 f are optimized by minimum-error training .	these models can be tuned using minimum error rate training .
xiao et al present a topic similarity model based on lda that produces a feature that weights grammar rules based on topic compatibility .	based on topic models , xiao et al present a topic similarity model for hpb system , where each rule is assigned with a topic distribution .
morphologically , arabic is a non-concatenative language .	moreover , arabic is a morphologically complex language .
text categorization is the classification of documents with respect to a set of predefined categories .	text categorization is a crucial and well-proven method for organizing the collection of large scale documents .
active learning processin this work , we are interested in selective sampling for pool-based active learning , and focus on uncertainty sampling .	in this work , we are interested in uncertainty sampling for pool-based active learning , in which an unlabeled example x with maximum uncertainty is selected for human annotation at each learning cycle .
word alignment is a crucial early step in the training of most statistical machine translation ( smt ) systems , in which the estimated alignments are used for constraining the set of candidates in phrase/grammar extraction ( cite-p-9-3-5 , cite-p-9-1-4 , cite-p-9-3-0 ) .	word alignment , which can be defined as an object for indicating the corresponding words in a parallel text , was first introduced as an intermediate result of statistical translation models ( cite-p-13-1-2 ) .
semantic parsing is the task of mapping natural language utterances to machine interpretable meaning representations .	semantic parsing is the mapping of text to a meaning representation .
first of all , multiple source languages can be involved to increase the statistical basis for learning , a strategy that can also be used in the case of annotation projection .	one idea is to use multiple source languages to increase the statistical ground for the learning process , a strategy that can also be used in the case of annotation projection .
for our experiments reported here , we obtained word vectors using the word2vec tool and the text8 corpus .	we learn our word embeddings by using word2vec 3 on unlabeled review data .
a 4-gram language model was trained on the target side of the parallel data using the srilm toolkit from stolcke .	the pre-processed monolingual sentences will be used by srilm or berkeleylm to train a n-gram language model .
we split each document into sentences using the sentence tokenizer of the nltk toolkit .	for this step we used regular expressions and nltk to tokenize the text .
for example , figure 1 shows a part of the variation n-grams found in the german tiger corpus .	as an example for these hierarchical relationships , figure 1 shows a german noun phrase taken from the german tiger corpus .
relation extraction ( re ) is the task of identifying instances of relations , such as nationality ( person , country ) or place of birth ( person , location ) , in passages of natural text .	relation extraction is the task of finding relations between entities in text , which is useful for several tasks such as information extraction , summarization , and question answering ( cite-p-14-3-7 ) .
the combination of ck with this additional knowledge follows the best settings from wiegand and klakow .	following wiegand and klakow , this corpus is chosen as a training set .
for the vector of document , the th element of is closely related to the generation probability of based on the language model induced by document .	the value at each dimension of the vector is closely related to the generation probability based on the language model of the corresponding document .
cite-p-24-3-1 presented a conditional variational framework for generating specific responses .	cite-p-24-3-10 introduced latent responding factors to model multiple responding mechanisms .
the reranking parser of charniak and johnson was used to parse the bnc .	all of the english sentences were parsed using the charniak parser .
a 5-gram language model was created with the sri language modeling toolkit and trained using the gigaword corpus and english sentences from the parallel data .	for both languages , we used the srilm toolkit to train a 5-gram language model using all monolingual data provided .
in the seminal work by rubenstein and goodenough , similarity judgments were obtained from 51 test subjects on 65 noun pairs written on paper cards .	for english , rubenstein and goodenough obtained similarity judgements from 51 subjects on 65 noun pairs , a seminal study which was later replicated by miller and charles , and resnik .
mitchell et al were the first to employ distributional semantic models to predict neural activation in the human brain using data obtained via functional magentic resonance imaging .	mitchell et al were the first to demonstrate that distributional semantic models encode some of the patterns found in the fmri data .
psl is a new model of statistical relation learning and has been quickly applied to solve many nlp and other machine learning tasks in recent years .	psl is a new statistical relational learning method that has been applied to many nlp and other machine learning tasks in recent years .
daum茅 and jagarlamudi use contextual and string similarity to mine translations for oov words in a high resource language domain adaptation for a machine translation setting .	daum茅 and jagarlamudi , zhang and zong , and irvine et al use new-domain comparable corpora to mine translations for unseen words .
twitter sentiment classification , which identifies the sentiment polarity of short , informal tweets , has attracted increasing research interest ( cite-p-15-1-20 , cite-p-15-1-19 ) .	twitter sentiment classification has attracted increasing research interest in recent years ( cite-p-15-1-20 , cite-p-15-1-19 ) .
we used the char representation strategy proposed by ma and hovy where char embeddings are combined using a convolutional neural network .	in order to capture long-range syntactic information for accurate disambiguation in pre-parsing phase , we build a lstm-crf model inspired by the neural network proposed in ma and hovy .
in the remainder of the paper , section 2 describes the sentence planning task .	in the remainder of the paper , section 2 describes the sentence planning task in more detail .
for the mnist , timit , and cifar dataset , that the generalization gap is not due to overfitting or overtraining , but due to different generalization capabilities of the local minima .	keskar et al . observe for the mnist , timit , and cifar dataset , that the generalization gap is not due to overfitting or overtraining , but due to different generalization capabilities of the local minima the networks converge to .
for the support vector machine , we used svm-light .	as a classifier , we chose support vector machines .
we use the attention-based nmt model introduced by bahdanau et al as our text-only nmt baseline .	we implement the attention model introduced by bahdanau et al which was the main technique for a sequence decoding in the last few years .
boyd-graber et al integrate a model of random walks on the wordnet graph into an lda topic model to build an unsupervised word sense disambiguation system .	boyd-graber et al incorporate the synset structure in wordnet into lda for word sense disambiguation , where each topic is a random process defined over the synsets .
keyphrases also offers a programming framework for developing new extraction algorithms .	the uima-based architecture of dkpro keyphrases allows users to easily evaluate keyphrase extraction configurations .
we measure machine translation performance using the bleu metric .	we evaluate the translation quality using the case-sensitive bleu-4 metric .
linear combinations of word embedding vectors have been shown to correspond well to the semantic composition of the individual words .	it has previously been shown that word embeddings represent the contextualised lexical semantics of words .
finally , we extract the semantic phrase table from the augmented aligned corpora using the moses toolkit .	in order to do so , we use the moses statistical machine translation toolkit .
results revealed that morphological and spontaneous speech-based features have an essential role in distinguishing mci patients from healthy controls .	our results suggest that it is primarily morphological and speech-based features that help distinguish mci patients from healthy controls .
for input representation , we used glove word embeddings .	our word embeddings is initialized with 100-dimensional glove word embeddings .
dependency parse correction , attachments in an input parse tree are revised by selecting , for a given dependent , the best governor from within a small set of candidates .	dependencies in an input parse tree are revised by selecting , for a given dependent , the best governor from within a small set of candidates .
and proved to be helpful for both inexperienced and experienced users .	results showed that the system was effective for inexperienced and experienced users .
we also used word2vec to generate dense word vectors for all word types in our learning corpus .	for this experiment , we used word2vec on the same frwac corpus to obtain a dense matrix in which each word is represented by a numeric vector .
the system described in this paper is the grandchild of the first transition-based neural network dependency parser ( cite-p-22-3-1 ) , which was the university of geneva ’ s entry in the conll 2007 multilingual dependency parsing shared task ( cite-p-22-1-7 ) .	the system described in this paper is a combination of a feature-based hierarchical lexicon and word grammar with an extended two-level morphology .
the popular method is to regard word segmentation as a sequence labeling problems .	one mainstream method is regarding word segmentation task as a sequence labeling problem .
we used the malt parser to obtain source english dependency trees and the stanford parser for arabic .	we used the stanford parser to generate dependency trees of sentences .
in the loss-augmented setting , the need of finding the max-violating constraint has severely limited the expressivity of effective loss functions .	in this setting , loss functions need to be factorizable together with the feature representations for finding the max-violating constraints .
a sentiment lexicon is a list of words and phrases , such as “ excellent ” , “ awful ” and “ not bad ” , each of them is assigned with a positive or negative score reflecting its sentiment polarity and strength ( cite-p-18-3-8 ) .	sentiment lexicon is a set of words ( or phrases ) each of which is assigned with a sentiment polarity score .
with english gigaword corpus , we use the skip-gram model as implemented in word2vec 3 to induce embeddings .	we use the word2vec cbow model with a window size of 5 and a minimum frequency of 5 to generate 200-dimensional vectors .
bidirectional rnns capture dependencies from both directions , thus providing two different views of the same sentence .	bidirectional rnns capture dependencies from both directions , thus provide two different views of the same sentence .
underspecification is nowadays the standard approach to dealing with scope ambiguities in computational semantics .	underspecification is the standard approach to dealing with scope ambiguities in computational semantics .
in this paper , a novel language model , the binarized embedding language model ( belm ) is proposed to solve the problem .	the first contribution in this paper is that a novel language model , the binarized embedding language model ( belm ) is proposed to reduce the memory consumption .
we built a 5-gram language model from it with the sri language modeling toolkit .	for both languages , we used the srilm toolkit to train a 5-gram language model using all monolingual data provided .
we use glove 300-dimension embedding vectors pre-trained on 840 billion tokens of web data .	in a second baseline model , we also incorporate 300-dimensional glove word embeddings trained on wikipedia and the gigaword corpus .
for language model , we use a trigram language model trained with the srilm toolkit on the english side of the training corpus .	for improving the word alignment , we use the word-classes that are trained from a monolingual corpus using the srilm toolkit .
marcu and echihabi 2002 ) proposed a method to identify discourse relations between text segments using na茂ve bayes classifiers trained on a huge corpus .	marcu and echihabi proposed a method for cheap acquisition of training data for discourse relation sense prediction .
a trigram language model with modified kneser-ney discounting and interpolation was used as produced by the srilm toolkit .	the model was built using the srilm toolkit with backoff and good-turing smoothing .
word sense disambiguation ( wsd ) is a problem of finding the relevant clues in a surrounding context .	word sense disambiguation ( wsd ) is a task to identify the intended sense of a word based on its context .
only henrich and hinrichs enrich the output of morphological segmentation with information from germanet to disambiguate such structures .	only henrich and hinrichs enrich the output of morphological segmentation with information from the annotated compounds of germanet to disambiguate such structures .
sentence , our method utilizes dependency structures and japanese dependency constraints to determine the word order of a translation .	to resolve the problem of generating a grammatically incorrect sentence , our method uses dependency structures and japanese dependency constraints to determine the word order of a translation .
knight and marcu proposed a sentence compression method using a noisy-channel model .	we presented a maximum entropy model to extend the sentence compression methods described by knight and marcu .
word alignment is the task of identifying corresponding words in sentence pairs .	word alignment is the task of identifying translational relations between words in parallel corpora , in which a word at one language is usually translated into several words at the other language ( fertility model ) ( cite-p-18-1-0 ) .
in this paper , we present a latent variable model for one-shot dialogue response , and investigate what kinds of diversity .	in this paper , we present a latent variable model to generate responses to input utterances .
proposed discriminative models are capable of incorporating domain knowledge , by adding diverse and overlapping features .	we show that discriminative models outperform the existing generative models by incorporating diverse features .
our baseline system is an standard phrase-based smt system built with moses .	our phrase-based mt system is trained by moses with standard parameters settings .
much recent work on language generation has made use of discourse representations based on rhetorical structure theory .	many recent studies in natural language processing have paid attention to rhetorical structure theory , a method of structured description of text .
we trained a smt system on 10k french-english sentences from the europarl corpus .	we trained the syntax-based system on 751,088 german-english translations from the europarl corpus .
sentence compression is the task of compressing long , verbose sentences into short , concise ones .	sentence compression is the task of compressing long sentences into short and concise ones by deleting words .
for smt decoding , we use the moses toolkit with kenlm for language model queries .	we use kenlm 3 for computing the target language model score .
tang et al proposed a user-product neural network to incorporate both user and product information for sentiment classification .	tang et al was first to incorporate user and product information into a neural network model for personalized rating prediction of products .
we used the kenlm language model toolkit with character 7-grams .	we used 4-gram language models , trained using kenlm .
we use 300 dimensional glove embeddings trained on the common crawl 840b tokens dataset , which remain fixed during training .	we use 300-dimensional glove vectors trained on 6b common crawl corpus as word embeddings , setting the embeddings of outof-vocabulary words to zero .
by simplifying the previously-proposed instance-based evaluation framework we are able to take advantage of crowdsourcing services .	our framework simplifies a previously proposed “ instance-based evaluation ” method that involved substantial annotator training , making it suitable for crowdsourcing .
we use srilm toolkit to train a trigram language model with modified kneser-ney smoothing on the target side of training corpus .	we use srilm for training a trigram language model on the english side of the training corpus .
conditional random fields are undirected graphical models trained to maximize the conditional probability of the desired outputs given the corresponding inputs .	conditional random fields are discriminatively-trained undirected graphical models that find the globally optimal labeling for a given configuration of random variables .
we used a 5-gram language model with modified kneser-ney smoothing , built with the srilm toolkit .	firstly , we built a forward 5-gram language model using the srilm toolkit with modified kneser-ney smoothing .
semantic role labeling ( srl ) is the task of labeling predicate-argument structure in sentences with shallow semantic information .	semantic role labeling ( srl ) has been defined as a sentence-level natural-language processing task in which semantic roles are assigned to the syntactic arguments of a predicate ( cite-p-14-1-7 ) .
the reordering model was trained with the hierarchical , monotone , swap , left to right bidirectional method and conditioned on both the source and target language .	the reordering model was trained with the hierarchical , monotone , swap , left to right bidirectional method and conditioned on both source and target language .
in this paper , novel syntactic sub-kernels are generated from the generalized kernel for the task of relation extraction .	using the generalized kernel , we will also propose a number of novel syntactic sub-kernels for relation extraction .
our baseline system is an standard phrase-based smt system built with moses .	our baseline is an in-house phrase-based statistical machine translation system very similar to moses .
a 4-gram language model was trained on the target side of the parallel data using the srilm toolkit from stolcke .	the language model was constructed using the srilm toolkit with interpolated kneser-ney discounting .
we use the moses software to train a pbmt model .	we use moses , an open source toolkit for training different systems .
wei and gulla modeled the hierarchical relation between product aspects .	wei and gulla , 2010 ) modeled the hierarchical relation between product aspects .
with the user and product attention , our model can take account of the global user preference and product characteristics .	with the user and product attention , our model can take account of the global user preference and product characteristics in both word level and semantic level .
other authors also report better results by using p-grams with the length in a range , rather than using p-grams of fixed length .	other authors also report better results by using n-grams with the length in a range , rather than using n-grams of fixed length .
in our study , we build a conditional probability model which will be described in detail .	in our study , we build a conditional probability model which will be described in detail in section 3.2.1 .
we use treetagger with the default parameter file for tokenization , lemmatization and annotation of part-of-speech information in the corpus .	we extract the part-of-speech tags for both source and translation sentences using treetagger .
we incorporate active learning into the biomedical named-entity recognition system to enhance the system ' s performance .	we incorporated mmr-based active machine-learning idea into the biomedical named-entity recognition system .
we used word2vec , a powerful continuous bag-of-words model to train word similarity .	we trained a continuous bag of words model of 400 dimensions and window size 5 with word2vec on the wiki set .
experimental results on four typical attributes showed that wikicike significantly outperforms both the current translation based methods and the monolingual extraction methods .	our experimental results demonstrate that wikicike outperforms the monolingual knowledge extraction method and the translation-based method .
we present the first neural endto-end solutions to computational am .	we present the first study on neural endto-end am .
in multimodal semantics , we evaluate on well known conceptual similarity and relatedness tasks and on zero-shot learning .	we use standard evaluations for multimodal semantics , including measuring conceptual similarity and cross-modal zero-shot learning .
word sense disambiguation ( wsd ) is the task of determining the meaning of a word in a given context .	word sense disambiguation ( wsd ) is a key enabling-technology that automatically chooses the intended sense of a word in context .
studies have also shown that the learned embedding captures both syntactic and semantic functions of words .	it has been empirically shown that word embeddings can capture semantic and syntactic similarities between words .
zoph et al train a parent model on a highresource language pair in order to improve low-resource language pairs .	zoph et al use multilingual transfer learning to improve nmt for lowresource languages .
compressing deep models into smaller networks has been an active area of research .	compressing deep learning models is an active area of current research .

in this paper , we examine user adaptation to the system ’ s lexical and syntactic choices in the context of the deployed .

we confirm prior results showing that users adapt to the system ’ s lexical and syntactic choices .

adwords features can be used for defining message persuasiveness .

adwords gives us an appropriate context for evaluating persuasive messages .

for faster training , we employ an efficient parallel training strategy proposed by mcdonald et al .

we make use of a distributed training strategy for the structured perceptron that was first introduced in mcdonald et al .

toutanova and moore addressed the phonetic substitution problem by extending the initial letter-to-phone model .

toutanova and moore improved this approach by extending the error model with phonetic similarities over words .

but the neural networks leverage these features for improving tagging .

this observation is evidence that the neural network can find good representations for pos tagging .

in general , the use of modifier structures and the associated semantic interpretation component permits a good treatment of scoping problems involving coordination .

the system 's semantic interpretation component can in particular deal with scoping problems involving coordination .

relation extraction is the task of predicting attributes and relations for entities in a sentence ( zelenko et al. , 2003 ; bunescu and mooney , 2005 ; guodong et al. , 2005 ) .

relation extraction ( re ) is the task of extracting semantic relationships between entities in text .

we describe an endto-end generation model that performs content selection and surface realization .

similar to cite-p-11-1-0 , we also present an endto-end system that performs content selection and surface realization .

sentiment detection and classification has received considerable attention .

sentiment classification has seen a great deal of attention .

the 位 f are optimized by minimum-error training .

these models can be tuned using minimum error rate training .

xiao et al present a topic similarity model based on lda that produces a feature that weights grammar rules based on topic compatibility .

based on topic models , xiao et al present a topic similarity model for hpb system , where each rule is assigned with a topic distribution .

morphologically , arabic is a non-concatenative language .

moreover , arabic is a morphologically complex language .

text categorization is the classification of documents with respect to a set of predefined categories .

text categorization is a crucial and well-proven method for organizing the collection of large scale documents .

active learning processin this work , we are interested in selective sampling for pool-based active learning , and focus on uncertainty sampling .

in this work , we are interested in uncertainty sampling for pool-based active learning , in which an unlabeled example x with maximum uncertainty is selected for human annotation at each learning cycle .

word alignment is a crucial early step in the training of most statistical machine translation ( smt ) systems , in which the estimated alignments are used for constraining the set of candidates in phrase/grammar extraction ( cite-p-9-3-5 , cite-p-9-1-4 , cite-p-9-3-0 ) .

word alignment , which can be defined as an object for indicating the corresponding words in a parallel text , was first introduced as an intermediate result of statistical translation models ( cite-p-13-1-2 ) .

semantic parsing is the task of mapping natural language utterances to machine interpretable meaning representations .

semantic parsing is the mapping of text to a meaning representation .

first of all , multiple source languages can be involved to increase the statistical basis for learning , a strategy that can also be used in the case of annotation projection .

one idea is to use multiple source languages to increase the statistical ground for the learning process , a strategy that can also be used in the case of annotation projection .

for our experiments reported here , we obtained word vectors using the word2vec tool and the text8 corpus .

we learn our word embeddings by using word2vec 3 on unlabeled review data .

a 4-gram language model was trained on the target side of the parallel data using the srilm toolkit from stolcke .

the pre-processed monolingual sentences will be used by srilm or berkeleylm to train a n-gram language model .

we split each document into sentences using the sentence tokenizer of the nltk toolkit .

for this step we used regular expressions and nltk to tokenize the text .

for example , figure 1 shows a part of the variation n-grams found in the german tiger corpus .

as an example for these hierarchical relationships , figure 1 shows a german noun phrase taken from the german tiger corpus .

relation extraction ( re ) is the task of identifying instances of relations , such as nationality ( person , country ) or place of birth ( person , location ) , in passages of natural text .

relation extraction is the task of finding relations between entities in text , which is useful for several tasks such as information extraction , summarization , and question answering ( cite-p-14-3-7 ) .

the combination of ck with this additional knowledge follows the best settings from wiegand and klakow .

following wiegand and klakow , this corpus is chosen as a training set .

for the vector of document , the th element of is closely related to the generation probability of based on the language model induced by document .

the value at each dimension of the vector is closely related to the generation probability based on the language model of the corresponding document .

cite-p-24-3-1 presented a conditional variational framework for generating specific responses .

cite-p-24-3-10 introduced latent responding factors to model multiple responding mechanisms .

the reranking parser of charniak and johnson was used to parse the bnc .

all of the english sentences were parsed using the charniak parser .

a 5-gram language model was created with the sri language modeling toolkit and trained using the gigaword corpus and english sentences from the parallel data .

for both languages , we used the srilm toolkit to train a 5-gram language model using all monolingual data provided .

in the seminal work by rubenstein and goodenough , similarity judgments were obtained from 51 test subjects on 65 noun pairs written on paper cards .

for english , rubenstein and goodenough obtained similarity judgements from 51 subjects on 65 noun pairs , a seminal study which was later replicated by miller and charles , and resnik .

mitchell et al were the first to employ distributional semantic models to predict neural activation in the human brain using data obtained via functional magentic resonance imaging .

mitchell et al were the first to demonstrate that distributional semantic models encode some of the patterns found in the fmri data .

psl is a new model of statistical relation learning and has been quickly applied to solve many nlp and other machine learning tasks in recent years .

psl is a new statistical relational learning method that has been applied to many nlp and other machine learning tasks in recent years .

daum茅 and jagarlamudi use contextual and string similarity to mine translations for oov words in a high resource language domain adaptation for a machine translation setting .

daum茅 and jagarlamudi , zhang and zong , and irvine et al use new-domain comparable corpora to mine translations for unseen words .

twitter sentiment classification , which identifies the sentiment polarity of short , informal tweets , has attracted increasing research interest ( cite-p-15-1-20 , cite-p-15-1-19 ) .

twitter sentiment classification has attracted increasing research interest in recent years ( cite-p-15-1-20 , cite-p-15-1-19 ) .

we used the char representation strategy proposed by ma and hovy where char embeddings are combined using a convolutional neural network .

in order to capture long-range syntactic information for accurate disambiguation in pre-parsing phase , we build a lstm-crf model inspired by the neural network proposed in ma and hovy .

in the remainder of the paper , section 2 describes the sentence planning task .

in the remainder of the paper , section 2 describes the sentence planning task in more detail .

for the mnist , timit , and cifar dataset , that the generalization gap is not due to overfitting or overtraining , but due to different generalization capabilities of the local minima .

keskar et al . observe for the mnist , timit , and cifar dataset , that the generalization gap is not due to overfitting or overtraining , but due to different generalization capabilities of the local minima the networks converge to .

for the support vector machine , we used svm-light .

as a classifier , we chose support vector machines .

we use the attention-based nmt model introduced by bahdanau et al as our text-only nmt baseline .

we implement the attention model introduced by bahdanau et al which was the main technique for a sequence decoding in the last few years .

boyd-graber et al integrate a model of random walks on the wordnet graph into an lda topic model to build an unsupervised word sense disambiguation system .

boyd-graber et al incorporate the synset structure in wordnet into lda for word sense disambiguation , where each topic is a random process defined over the synsets .

keyphrases also offers a programming framework for developing new extraction algorithms .

the uima-based architecture of dkpro keyphrases allows users to easily evaluate keyphrase extraction configurations .

we measure machine translation performance using the bleu metric .

we evaluate the translation quality using the case-sensitive bleu-4 metric .

linear combinations of word embedding vectors have been shown to correspond well to the semantic composition of the individual words .

it has previously been shown that word embeddings represent the contextualised lexical semantics of words .

finally , we extract the semantic phrase table from the augmented aligned corpora using the moses toolkit .

in order to do so , we use the moses statistical machine translation toolkit .

results revealed that morphological and spontaneous speech-based features have an essential role in distinguishing mci patients from healthy controls .

our results suggest that it is primarily morphological and speech-based features that help distinguish mci patients from healthy controls .

for input representation , we used glove word embeddings .

our word embeddings is initialized with 100-dimensional glove word embeddings .

dependency parse correction , attachments in an input parse tree are revised by selecting , for a given dependent , the best governor from within a small set of candidates .

dependencies in an input parse tree are revised by selecting , for a given dependent , the best governor from within a small set of candidates .

and proved to be helpful for both inexperienced and experienced users .

results showed that the system was effective for inexperienced and experienced users .

we also used word2vec to generate dense word vectors for all word types in our learning corpus .

for this experiment , we used word2vec on the same frwac corpus to obtain a dense matrix in which each word is represented by a numeric vector .

the system described in this paper is the grandchild of the first transition-based neural network dependency parser ( cite-p-22-3-1 ) , which was the university of geneva ’ s entry in the conll 2007 multilingual dependency parsing shared task ( cite-p-22-1-7 ) .

the system described in this paper is a combination of a feature-based hierarchical lexicon and word grammar with an extended two-level morphology .

the popular method is to regard word segmentation as a sequence labeling problems .

one mainstream method is regarding word segmentation task as a sequence labeling problem .

we used the malt parser to obtain source english dependency trees and the stanford parser for arabic .

we used the stanford parser to generate dependency trees of sentences .

in the loss-augmented setting , the need of finding the max-violating constraint has severely limited the expressivity of effective loss functions .

in this setting , loss functions need to be factorizable together with the feature representations for finding the max-violating constraints .

a sentiment lexicon is a list of words and phrases , such as “ excellent ” , “ awful ” and “ not bad ” , each of them is assigned with a positive or negative score reflecting its sentiment polarity and strength ( cite-p-18-3-8 ) .

sentiment lexicon is a set of words ( or phrases ) each of which is assigned with a sentiment polarity score .

with english gigaword corpus , we use the skip-gram model as implemented in word2vec 3 to induce embeddings .

we use the word2vec cbow model with a window size of 5 and a minimum frequency of 5 to generate 200-dimensional vectors .

bidirectional rnns capture dependencies from both directions , thus providing two different views of the same sentence .

bidirectional rnns capture dependencies from both directions , thus provide two different views of the same sentence .

underspecification is nowadays the standard approach to dealing with scope ambiguities in computational semantics .

underspecification is the standard approach to dealing with scope ambiguities in computational semantics .

in this paper , a novel language model , the binarized embedding language model ( belm ) is proposed to solve the problem .

the first contribution in this paper is that a novel language model , the binarized embedding language model ( belm ) is proposed to reduce the memory consumption .

we built a 5-gram language model from it with the sri language modeling toolkit .

for both languages , we used the srilm toolkit to train a 5-gram language model using all monolingual data provided .

we use glove 300-dimension embedding vectors pre-trained on 840 billion tokens of web data .

in a second baseline model , we also incorporate 300-dimensional glove word embeddings trained on wikipedia and the gigaword corpus .

for language model , we use a trigram language model trained with the srilm toolkit on the english side of the training corpus .

for improving the word alignment , we use the word-classes that are trained from a monolingual corpus using the srilm toolkit .

marcu and echihabi 2002 ) proposed a method to identify discourse relations between text segments using na茂ve bayes classifiers trained on a huge corpus .

marcu and echihabi proposed a method for cheap acquisition of training data for discourse relation sense prediction .

a trigram language model with modified kneser-ney discounting and interpolation was used as produced by the srilm toolkit .

the model was built using the srilm toolkit with backoff and good-turing smoothing .

word sense disambiguation ( wsd ) is a problem of finding the relevant clues in a surrounding context .

word sense disambiguation ( wsd ) is a task to identify the intended sense of a word based on its context .

only henrich and hinrichs enrich the output of morphological segmentation with information from germanet to disambiguate such structures .

only henrich and hinrichs enrich the output of morphological segmentation with information from the annotated compounds of germanet to disambiguate such structures .

sentence , our method utilizes dependency structures and japanese dependency constraints to determine the word order of a translation .

to resolve the problem of generating a grammatically incorrect sentence , our method uses dependency structures and japanese dependency constraints to determine the word order of a translation .

knight and marcu proposed a sentence compression method using a noisy-channel model .

we presented a maximum entropy model to extend the sentence compression methods described by knight and marcu .

word alignment is the task of identifying corresponding words in sentence pairs .

word alignment is the task of identifying translational relations between words in parallel corpora , in which a word at one language is usually translated into several words at the other language ( fertility model ) ( cite-p-18-1-0 ) .

in this paper , we present a latent variable model for one-shot dialogue response , and investigate what kinds of diversity .

in this paper , we present a latent variable model to generate responses to input utterances .

proposed discriminative models are capable of incorporating domain knowledge , by adding diverse and overlapping features .

we show that discriminative models outperform the existing generative models by incorporating diverse features .

our baseline system is an standard phrase-based smt system built with moses .

our phrase-based mt system is trained by moses with standard parameters settings .

much recent work on language generation has made use of discourse representations based on rhetorical structure theory .

many recent studies in natural language processing have paid attention to rhetorical structure theory , a method of structured description of text .

we trained a smt system on 10k french-english sentences from the europarl corpus .

we trained the syntax-based system on 751,088 german-english translations from the europarl corpus .

sentence compression is the task of compressing long , verbose sentences into short , concise ones .

sentence compression is the task of compressing long sentences into short and concise ones by deleting words .

for smt decoding , we use the moses toolkit with kenlm for language model queries .

we use kenlm 3 for computing the target language model score .

tang et al proposed a user-product neural network to incorporate both user and product information for sentiment classification .

tang et al was first to incorporate user and product information into a neural network model for personalized rating prediction of products .

we used the kenlm language model toolkit with character 7-grams .

we used 4-gram language models , trained using kenlm .

we use 300 dimensional glove embeddings trained on the common crawl 840b tokens dataset , which remain fixed during training .

we use 300-dimensional glove vectors trained on 6b common crawl corpus as word embeddings , setting the embeddings of outof-vocabulary words to zero .

by simplifying the previously-proposed instance-based evaluation framework we are able to take advantage of crowdsourcing services .

our framework simplifies a previously proposed “ instance-based evaluation ” method that involved substantial annotator training , making it suitable for crowdsourcing .

we use srilm toolkit to train a trigram language model with modified kneser-ney smoothing on the target side of training corpus .

we use srilm for training a trigram language model on the english side of the training corpus .

conditional random fields are undirected graphical models trained to maximize the conditional probability of the desired outputs given the corresponding inputs .

conditional random fields are discriminatively-trained undirected graphical models that find the globally optimal labeling for a given configuration of random variables .

we used a 5-gram language model with modified kneser-ney smoothing , built with the srilm toolkit .

firstly , we built a forward 5-gram language model using the srilm toolkit with modified kneser-ney smoothing .

semantic role labeling ( srl ) is the task of labeling predicate-argument structure in sentences with shallow semantic information .

semantic role labeling ( srl ) has been defined as a sentence-level natural-language processing task in which semantic roles are assigned to the syntactic arguments of a predicate ( cite-p-14-1-7 ) .

the reordering model was trained with the hierarchical , monotone , swap , left to right bidirectional method and conditioned on both the source and target language .

the reordering model was trained with the hierarchical , monotone , swap , left to right bidirectional method and conditioned on both source and target language .

in this paper , novel syntactic sub-kernels are generated from the generalized kernel for the task of relation extraction .

using the generalized kernel , we will also propose a number of novel syntactic sub-kernels for relation extraction .

our baseline system is an standard phrase-based smt system built with moses .

our baseline is an in-house phrase-based statistical machine translation system very similar to moses .

a 4-gram language model was trained on the target side of the parallel data using the srilm toolkit from stolcke .

the language model was constructed using the srilm toolkit with interpolated kneser-ney discounting .

we use the moses software to train a pbmt model .

we use moses , an open source toolkit for training different systems .

wei and gulla modeled the hierarchical relation between product aspects .

wei and gulla , 2010 ) modeled the hierarchical relation between product aspects .

with the user and product attention , our model can take account of the global user preference and product characteristics .

with the user and product attention , our model can take account of the global user preference and product characteristics in both word level and semantic level .

other authors also report better results by using p-grams with the length in a range , rather than using p-grams of fixed length .

other authors also report better results by using n-grams with the length in a range , rather than using n-grams of fixed length .

in our study , we build a conditional probability model which will be described in detail .

in our study , we build a conditional probability model which will be described in detail in section 3.2.1 .

we use treetagger with the default parameter file for tokenization , lemmatization and annotation of part-of-speech information in the corpus .

we extract the part-of-speech tags for both source and translation sentences using treetagger .

we incorporate active learning into the biomedical named-entity recognition system to enhance the system ' s performance .

we incorporated mmr-based active machine-learning idea into the biomedical named-entity recognition system .

we used word2vec , a powerful continuous bag-of-words model to train word similarity .

we trained a continuous bag of words model of 400 dimensions and window size 5 with word2vec on the wiki set .

experimental results on four typical attributes showed that wikicike significantly outperforms both the current translation based methods and the monolingual extraction methods .

our experimental results demonstrate that wikicike outperforms the monolingual knowledge extraction method and the translation-based method .

we present the first neural endto-end solutions to computational am .

we present the first study on neural endto-end am .

in multimodal semantics , we evaluate on well known conceptual similarity and relatedness tasks and on zero-shot learning .

we use standard evaluations for multimodal semantics , including measuring conceptual similarity and cross-modal zero-shot learning .

word sense disambiguation ( wsd ) is the task of determining the meaning of a word in a given context .

word sense disambiguation ( wsd ) is a key enabling-technology that automatically chooses the intended sense of a word in context .

studies have also shown that the learned embedding captures both syntactic and semantic functions of words .

it has been empirically shown that word embeddings can capture semantic and syntactic similarities between words .

zoph et al train a parent model on a highresource language pair in order to improve low-resource language pairs .

zoph et al use multilingual transfer learning to improve nmt for lowresource languages .

compressing deep models into smaller networks has been an active area of research .

compressing deep learning models is an active area of current research .