HHousen/ParaSCI · Datasets at Hugging Face

sentence1 stringlengths 16 446	sentence2 stringlengths 14 436
the 5-gram kneser-ney smoothed language models were trained by srilm , with kenlm used at runtime .	the language model was a kneser-ney interpolated trigram model generated using the srilm toolkit .
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 recognizing and extracting relations between entities or concepts in texts .
a 4-gram language model was trained on the target side of the parallel data using the srilm toolkit from stolcke .	a trigram model was built on 20 million words of general newswire text , using the srilm toolkit .
we then made videos of every schedule for every sentence , using the festival speech synthesiser and the ruth talking head .	videos were then created of all of the schedules for all of the sentences , using the festival speech synthesiser and the ruth animated talking head .
in this paper , we proposed a novel algorithm , show-and-fool , for crafting adversarial examples .	in this paper , we tackle the aforementioned challenges by proposing a novel algorithm called show-and-fool .
the long short-term memory was first proposed by hochreiter and schmidhuber that can learn long-term dependencies .	long short-term memory was introduced by hochreiter and schmidhuber to overcome the issue of vanishing gradients in the vanilla recurrent neural networks .
nlp tasks are publicly available : datasets for word segmentation and pos tagging were released for the first vlsp evaluation campaign .	vncorenlp provides core nlp steps including word segmentation , pos tagging , ner and dependency parsing .
over multi-domain language identification and multi-domain sentiment analysis , we show our models to substantially outperform a baseline deep learning method , and set a new benchmark for state-of-the-art cross-domain .	evaluating on multi-domain language identification and multi-domain sentiment analysis , we show substantial improvements over standard domain adaptation techniques , and domain-adversarial training .
the character embeddings are computed using a method similar to word2vec .	the embedding layer was initialized using word2vec vectors .
for estimating monolingual word vector models , we use the cbow algorithm as implemented in the word2vec package using a 5-token window .	as monolingual baselines , we use the skip-gram and cbow methods of mikolov et al as implemented in the gensim package .
we use the standard generative dependency model with valence .	our experiments use the dependency model with valence .
in each plot , a single arrow signifies one word , pointing from the position of the original word .	in each plot , a single arrow signifies one word , pointing from the position of the original word embedding to the updated representation .
trigram language models are implemented using the srilm toolkit .	srilm toolkit is used to build these language models .
this is partly in line with sahlgren and lenci who observed that it is more challenging for neural-based models to train good vectors for low-frequency words .	this is in line with sahlgren and lenci who showed that dsms perform the best for medium to high-frequency ranges items .
the nodes are concepts ( or synsets as they are called in the wordnet ) .	wordnet is a general english thesaurus which additionally covers biological terms .
we use case-sensitive bleu-4 to measure the quality of translation result .	we evaluate the translation quality using the case-sensitive bleu-4 metric .
for chinese , we exploit wikipedia documents to train the same dimensional word2vec embeddings .	with english gigaword corpus , we use the skip-gram model as implemented in word2vec 3 to induce embeddings .
the universal dependencies is a worldwide project to provide multilingual syntactic resources of dependency structures with a uniformed tag set for all languages .	the universal dependencies project is a recent effort aimed at facilitating crosslingual parsing development through the standardization of dependency annotation schemes across languages .
in addition , instead of using the popular crf model , we use another sequence labeling model in this paper -- -the hidden markov support vector machines model .	instead of using crf model , we use the hidden markov support vector machines , which is also a sequence labeling model like crf .
additionally , we compile the model using the adamax optimizer .	we use the adam optimizer and mini-batch gradient to solve this optimization problem .
we used the pharaoh decoder for both the minimum error rate training and test dataset decoding .	we used minimum error rate training to optimize the feature weights .
we first create a sentence quotation graph to represent the conversation structure .	we first build a sentence quotation graph that captures the conversation structure among emails .
cite-p-17-1-3 used an lstm architecture to capture potential long-distance dependencies , which alleviates the limitation of the size of context window .	cite-p-17-5-5 used a linear-time incremental model which can also benefits from various kinds of features including word-based features .
in this paper , we present a reinforcement learning framework for inducing mappings from text to actions .	in this paper , we presented a reinforcement learning approach for inducing a mapping between instructions and actions .
b ing produces a much better translation : chef d ¡¯ etat-major de la defense du mali .	then b ing produces a much better translation : chef d¡¯¨¦tat-major de la d¨¦fense du mali veut plus d¡¯armes .
in this paper , we propose another phrase-level combination approach ¨c a paraphrasing model .	in this paper , we propose a paraphrasing model to address the task of system combination for machine translation .
coreference resolution is the process of finding discourse entities ( markables ) referring to the same real-world entity or concept .	coreference resolution is the problem of partitioning a sequence of noun phrases ( or mentions ) , as they occur in a natural language text , into a set of referential entities .
the weights of the different feature functions were optimised by means of minimum error rate training on the 2008 test set .	the feature weights for each system were tuned on development sets using the moses implementation of minimum error rate training .
we use srilm toolkit to train a trigram language model with modified kneser-ney smoothing on the target side of training corpus .	we train a 4-gram language model on the xinhua portion of the english gigaword corpus using the srilm toolkits with modified kneser-ney smoothing .
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 detecting and characterizing semantic relations between entities from free text .
the character embeddings are computed using a method similar to word2vec .	the words of input sentences are first converted to vector representations learned from word2vec tool .
we used moses , a phrase-based smt toolkit , for training the translation model .	we adapted the moses phrase-based decoder to translate word lattices .
in a naive implementation , a new phrase t . vpe is built by copying older ones and then combining the copies according to the constraints stated in a grammar .	in a naive implementation , a new phrase type is built by copying older ones and then combining the copies according to the constraints stated in a grammar rule .
to deal with this problem , we propose graph merging , a new perspective , for building flexible dependency graphs .	to deal with this problem , we propose graph merging , a new perspective , for building flexible representations .
our baseline is a phrase-based mt system trained using the moses toolkit .	we use the moses toolkit to train our phrase-based smt models .
we use the pre-trained glove 50-dimensional word embeddings to represent words found in the glove dataset .	we use 300 dimensional glove embeddings trained on the common crawl 840b tokens dataset , which remain fixed during training .
we used the pre-trained word embeddings that are learned using the word2vec toolkit on google news dataset .	we use publicly-available 1 300-dimensional embeddings trained on part of the google news dataset using skip-gram with negative sampling .
socher et al proposed the recursive neural network that has been proven to be efficient in terms of constructing sentences representations .	socher et al introduce a family of recursive neural networks for sentence-level semantic composition .
xiong et al develop a bottom-up decoder for btg that uses only phrase pairs .	feng et al use shift-reduce parsing to impose itg constraints on phrase permutation .
we use our implementation of hierarchical phrase-based smt , with standard features , for the smt experiments .	in our experiments the mt system used is hierarchical phrase-based system .
word sense disambiguation ( wsd ) is the task of identifying the correct meaning of a word in context .	word sense disambiguation ( wsd ) is a key enabling-technology .
language modeling is trained using kenlm using 5-grams , with modified kneser-ney smoothing .	the language models are 4-grams with modified kneser-ney smoothing which have been trained with the srilm toolkit .
the dataset we used in the present study is the online edition 2 of the world atlas of language structures .	we use the world atlas of language structure dataset , on which we conduct experiments .
modi et al extended the model of , which is an unsupervised model for inducing semantic roles , to jointly induce semantic roles and frames across verbs using the chinese restaurant process .	modi et al extended the model of to jointly induce semantic roles and frames using the chinese restaurant process , which is also used in our approach .
information extraction ( ie ) is the task of identifying information in texts and converting it into a predefined format .	information extraction ( ie ) is a technology that can be applied to identifying both sources and targets of new hyperlinks .
coreference resolution is a fundamental component of natural language processing ( nlp ) and has been widely applied in other nlp tasks ( cite-p-15-3-9 ) .	since coreference resolution is a pervasive discourse phenomenon causing performance impediments in current ie systems , we considered a corpus of aligned english and romanian texts to identify coreferring expressions .
for the lda based method , adding other content words , combined with an increased number of topics , can further improve the performance , achieving up to 14 . 23 % perplexity reduction .	in addition , we find that for the lda based adaptation scheme , adding more content words and increasing the number of topics can further improve the performance significantly .
sentiment analysis in twitter , which is a task of semeval , was firstly proposed in 2013 and not replaced until 2018 .	sentiment analysis in twitter is a particularly challenging task , because of the informal and “ creative ” writing style , with improper use of grammar , figurative language , misspellings and slang .
output string is guaranteed to conform to a given target grammar .	output always conforms to the given target grammar .
in this paper , we exploit non-local features as an estimate of long-distance dependencies .	in this paper , our approach describes how to exploit non-local information to a slu problem .
the smt systems were trained using the moses toolkit and the experiment management system .	the experiment management system from the open source moses smt toolkit was used to conduct the experiments .
these connections may be derived from work in language assessment and grade expectations such as found in , and .	these connections may be derived from work in language assessment and grade expectations such as found in .
the probabilistic language model is constructed on google web 1t 5-gram corpus by using the srilm toolkit .	a standard sri 5-gram language model is estimated from monolingual data .
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 ) .	word alignment is a critical first step for building statistical machine translation systems .
transition-based methods have given competitive accuracies and efficiencies for dependency parsing .	transition-based and graph-based models have attracted the most attention of dependency parsing in recent years .
zhuang et al present an algorithm for the extraction of opinion target -opinion word pairs .	zhuang et al present a supervised algorithm for the extraction of opinion expression -opinion target pairs .
therefore , we adopt the greedy feature selection algorithm as described in jiang et al to pick up positive features incrementally according to their contributions .	we adopt the greedy feature selection algorithm as described in jiang and ng to pick up positive features empirically and incrementally according to their contributions on the development data .
we estimated 5-gram language models using the sri toolkit with modified kneser-ney smoothing .	further , we apply a 4-gram language model trained with the srilm toolkit on the target side of the training corpus .
gold has been created to facilitate a more standardised use of basic grammatical features .	the idea behind gold is to facilitate a more standardised use of basic grammatical features .
therefore , we use the long short-term memory network to overcome this problem .	an effective solution for these problems is the long short-term memory architecture .
as described in this paper , we demonstrate the contribution of modality analysis for disease .	in this study , we use generic modality features to improve factuality analysis .
the first dataset is mov , which is a classical movie review dataset .	the first dataset is mov , which is a widely-used movie review dataset .
crfs have been shown to perform well on a number of nlp problems such as shallow parsing , table extraction , and named entity recognition .	crfs have been shown to perform well in a number of natural language processing applications , such as pos tagging , shallow parsing or np chunking , and named entity recognition .
ji and grishman employ an approach to propagate consistent event arguments across sentences and documents .	later , ji and grishman employed a rule-based approach to propagate consistent triggers and arguments across topic-related documents .
for training our system classifier , we have used scikit-learn .	we implement classification models using keras and scikit-learn .
the language models were built using srilm toolkits .	the srilm toolkit was used to build the 5-gram language model .
these embeddings are determined beforehand on a very large corpus typically using either the skip gram or the continuous bag of words variant of the word2vec model .	these word embeddings are learned in advance using a continuous skip-gram model , or other continuous word representation learning methods .
quirk et al extended path to treelets and put forward dependency treelet translation .	quirk et al and xiong et al used treelets to model the source dependency tree using synchronous grammars .
the spelling correction models from brill and moore and toutanova and moore use the noisy channel model approach to determine the types and weights of edit operations .	the spelling error model proposed by brill and moore allows generic string edit operations up to a certain length .
we use the sri language model toolkit to train a 5-gram model with modified kneser-ney smoothing on the target-side training corpus .	gram language model with modified kneser-ney smoothing is trained with the srilm toolkit on the epps , ted , newscommentary , and the gigaword corpora .
this is the first work that applies a state-of-the-art probabilistic parsing model to al for dependency parsing .	this paper for the first time applies a state-of-the-art probabilistic model to al with pa for dependency parsing .
the methods employed for gathering the data , preparation and compilation of dataset , used in offenseval shared task is described in zampieri et al .	the data collection methods used to compile the dataset used in offenseval is described in zampieri et al .
for evaluation we use mteval-v13a from the moses toolkit and tercom 3 to score our systems on the bleu respectively ter measures .	we use mteval from the moses toolkit an tercom to evaluate our systems on the bleu and ter measures .
bisk and hockenmaier used combinatory categorial grammar to learn syntactic dependencies from word strings .	bisk and hockenmaier use an em approach to induce a combinatory categorial grammar , based on very general linguistic assumptions .
the sg model is a popular choice to learn word embeddings by leveraging the relations between a word and its neighboring words .	the skip-gram model is a very popular technique for learning embeddings that scales to huge corpora and can capture important semantic and syntactic properties of words .
for word embeddings , we used popular pre-trained word vectors from glove .	we use the pre-trained glove vectors to initialize word embeddings .
wasserstein distance takes into account the cross-term relationship between different words in a principled fashion .	by using the wasserstein distance between distributions , the wordto-word semantic relationship is taken into account in a principled way .
eurowordnet is a multilingual lexical knowledge base comprised of hierarchical representations of lexical items for several european languages .	eurowordnet is a multilingual semantic lexicon with wordnets for several european languages , which are structured as the princeton wordnet .
klein and manning demonstrated that linguistically informed splitting of nonterminal symbols in treebank-derived grammars can result in accurate grammars .	klein and manning identified nonterminals that could valuably be split into fine-grained ones using hand-written linguistic rules .
sentiment classification is a well-studied and active research area ( cite-p-20-1-11 ) .	sentiment classification is a special task of text categorization that aims to classify documents according to their opinion of , or sentiment toward a given subject ( e.g. , if an opinion is supported or not ) ( cite-p-11-1-2 ) .
we will evaluate the performance of sampling distributions based on perplexities calculated using small , lightweight rnn language models .	we experimentally evaluate the heldout perplexity of models trained with our various importance sampling distributions .
for english , number and gender for common nouns are computed via a comparison of head lemma to head and using the number and gender data of bergsma and lin .	we compute number and gender for common nouns using the number and gender data provided by bergsma and lin .
a tree domain is a subset of strings over a linearly ordered set which is closed under prefix and left sister .	a tree domain is a set of node address drawn from n * ( that is , a set of strings of natural numbers ) in which c is the address of the root and the children of a node at address w occur at addresses w0 , wl , ... , in leftto-right order .
for our purpose we use word2vec embeddings trained on a google news dataset and find the pairwise cosine distances for all words .	we use the word2vec vectors with 300 dimensions , pre-trained on 100 billion words of google news .
ritchie et al used a combination of terms from citation contexts and existing index terms of a paper to improve indexing of cited papers .	for example , ritchie et al used a combination of terms from citation contexts and existing index terms of a paper to improve indexing of cited papers .
tag is a tree-rewriting system : the derivation process consists in applying operations to trees in order to obtain a ( derived ) tree whose sequence of leaves is a sentence .	a tag is a rewriting system that derives trees starting from a finite set of elementary trees .
we used implementations from scikitlearn , and the parameters of both classifiers were tuned on the development set using grid search .	we trained the three classifiers using the svm implementation in scikit-learn , and tuned hyper-parameters c and 纬 using 10-fold cross-validation with the train split .
dependency parsing is a simpler task than constituent parsing , since dependency trees do not have extra non-terminal nodes and there is no need for a grammar to generate them .	dependency parsing is the task of predicting the most probable dependency structure for a given sentence .
table 4 shows the bleu scores of the output descriptions .	the numbers in the table are bleu scores of different neural models .
we created 5-gram language models for every domain using srilm with improved kneserney smoothing on the target side of the training parallel corpora .	we train a 4-gram language model on the xinhua portion of the english gigaword corpus using the srilm toolkits with modified kneser-ney smoothing .
for parsing arabic texts into syntactic trees , we used the berkeley parser .	for our investigations , we used the berkeley parser as a source of grammar rule clusters .
in doing so we can achieve better word retrieval performance than language models with only n-gram context .	we aim to improve speech retrieval performance by augmenting traditional n-gram language models with different types of topic context .
svmhmms and crfs have been successfully applied to a range of sequential tagging tasks such as syllabification , chunk parsing and word segmentation .	crfs have been shown to perform well in a number of natural language processing applications , such as pos tagging , shallow parsing or np chunking , and named entity recognition .
in this paper , we name the problem of choosing the correct word from the homophone set .	in this paper , we used the decision list to solve the homophone problem .
we use the max-loss variant of the margin infused relaxed algorithm with a hamming-loss margin as is common in the dependency parsing literature .	we select the cutting-plane variant of the margin-infused relaxed algorithm with additional extensions described by eidelman .
we use a baseline parser to parse large-scale unannotated data .	first , we use a baseline parser to parse large-scale unannotated data .
system tuning was carried out using both k-best mira and minimum error rate training on the held-out development set .	feature weights were set with minimum error rate training on a development set using bleu as the objective function .
with the rise of social media , more and more user generated sentiment data have been shared on the web .	the rise of social media such as blogs and microblogs has fueled interest in sentiment analysis .
as a classifier , we choose a first-order conditional random field model .	to this end , we use first-and second-order conditional random fields .
mauser et al extended this model to condition it on source word cooccurrences .	mauser et al integrated a logistic regression model predicting target words from all the source words in a pbsmt .

the 5-gram kneser-ney smoothed language models were trained by srilm , with kenlm used at runtime .

the language model was a kneser-ney interpolated trigram model generated using the srilm toolkit .

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 recognizing and extracting relations between entities or concepts in texts .

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

a trigram model was built on 20 million words of general newswire text , using the srilm toolkit .

we then made videos of every schedule for every sentence , using the festival speech synthesiser and the ruth talking head .

videos were then created of all of the schedules for all of the sentences , using the festival speech synthesiser and the ruth animated talking head .

in this paper , we proposed a novel algorithm , show-and-fool , for crafting adversarial examples .

in this paper , we tackle the aforementioned challenges by proposing a novel algorithm called show-and-fool .

the long short-term memory was first proposed by hochreiter and schmidhuber that can learn long-term dependencies .

long short-term memory was introduced by hochreiter and schmidhuber to overcome the issue of vanishing gradients in the vanilla recurrent neural networks .

nlp tasks are publicly available : datasets for word segmentation and pos tagging were released for the first vlsp evaluation campaign .

vncorenlp provides core nlp steps including word segmentation , pos tagging , ner and dependency parsing .

over multi-domain language identification and multi-domain sentiment analysis , we show our models to substantially outperform a baseline deep learning method , and set a new benchmark for state-of-the-art cross-domain .

evaluating on multi-domain language identification and multi-domain sentiment analysis , we show substantial improvements over standard domain adaptation techniques , and domain-adversarial training .

the character embeddings are computed using a method similar to word2vec .

the embedding layer was initialized using word2vec vectors .

for estimating monolingual word vector models , we use the cbow algorithm as implemented in the word2vec package using a 5-token window .

as monolingual baselines , we use the skip-gram and cbow methods of mikolov et al as implemented in the gensim package .

we use the standard generative dependency model with valence .

our experiments use the dependency model with valence .

in each plot , a single arrow signifies one word , pointing from the position of the original word .

in each plot , a single arrow signifies one word , pointing from the position of the original word embedding to the updated representation .

trigram language models are implemented using the srilm toolkit .

srilm toolkit is used to build these language models .

this is partly in line with sahlgren and lenci who observed that it is more challenging for neural-based models to train good vectors for low-frequency words .

this is in line with sahlgren and lenci who showed that dsms perform the best for medium to high-frequency ranges items .

the nodes are concepts ( or synsets as they are called in the wordnet ) .

wordnet is a general english thesaurus which additionally covers biological terms .

we use case-sensitive bleu-4 to measure the quality of translation result .

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

for chinese , we exploit wikipedia documents to train the same dimensional word2vec embeddings .

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

the universal dependencies is a worldwide project to provide multilingual syntactic resources of dependency structures with a uniformed tag set for all languages .

the universal dependencies project is a recent effort aimed at facilitating crosslingual parsing development through the standardization of dependency annotation schemes across languages .

in addition , instead of using the popular crf model , we use another sequence labeling model in this paper -- -the hidden markov support vector machines model .

instead of using crf model , we use the hidden markov support vector machines , which is also a sequence labeling model like crf .

additionally , we compile the model using the adamax optimizer .

we use the adam optimizer and mini-batch gradient to solve this optimization problem .

we used the pharaoh decoder for both the minimum error rate training and test dataset decoding .

we used minimum error rate training to optimize the feature weights .

we first create a sentence quotation graph to represent the conversation structure .

we first build a sentence quotation graph that captures the conversation structure among emails .

cite-p-17-1-3 used an lstm architecture to capture potential long-distance dependencies , which alleviates the limitation of the size of context window .

cite-p-17-5-5 used a linear-time incremental model which can also benefits from various kinds of features including word-based features .

in this paper , we present a reinforcement learning framework for inducing mappings from text to actions .

in this paper , we presented a reinforcement learning approach for inducing a mapping between instructions and actions .

b ing produces a much better translation : chef d ¡¯ etat-major de la defense du mali .

then b ing produces a much better translation : chef d¡¯¨¦tat-major de la d¨¦fense du mali veut plus d¡¯armes .

in this paper , we propose another phrase-level combination approach ¨c a paraphrasing model .

in this paper , we propose a paraphrasing model to address the task of system combination for machine translation .

coreference resolution is the process of finding discourse entities ( markables ) referring to the same real-world entity or concept .

coreference resolution is the problem of partitioning a sequence of noun phrases ( or mentions ) , as they occur in a natural language text , into a set of referential entities .

the weights of the different feature functions were optimised by means of minimum error rate training on the 2008 test set .

the feature weights for each system were tuned on development sets using the moses implementation of minimum error rate training .

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

we train a 4-gram language model on the xinhua portion of the english gigaword corpus using the srilm toolkits with modified kneser-ney smoothing .

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 detecting and characterizing semantic relations between entities from free text .

the character embeddings are computed using a method similar to word2vec .

the words of input sentences are first converted to vector representations learned from word2vec tool .

we used moses , a phrase-based smt toolkit , for training the translation model .

we adapted the moses phrase-based decoder to translate word lattices .

in a naive implementation , a new phrase t . vpe is built by copying older ones and then combining the copies according to the constraints stated in a grammar .

in a naive implementation , a new phrase type is built by copying older ones and then combining the copies according to the constraints stated in a grammar rule .

to deal with this problem , we propose graph merging , a new perspective , for building flexible dependency graphs .

to deal with this problem , we propose graph merging , a new perspective , for building flexible representations .

our baseline is a phrase-based mt system trained using the moses toolkit .

we use the moses toolkit to train our phrase-based smt models .

we use the pre-trained glove 50-dimensional word embeddings to represent words found in the glove dataset .

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

we used the pre-trained word embeddings that are learned using the word2vec toolkit on google news dataset .

we use publicly-available 1 300-dimensional embeddings trained on part of the google news dataset using skip-gram with negative sampling .

socher et al proposed the recursive neural network that has been proven to be efficient in terms of constructing sentences representations .

socher et al introduce a family of recursive neural networks for sentence-level semantic composition .

xiong et al develop a bottom-up decoder for btg that uses only phrase pairs .

feng et al use shift-reduce parsing to impose itg constraints on phrase permutation .

we use our implementation of hierarchical phrase-based smt , with standard features , for the smt experiments .

in our experiments the mt system used is hierarchical phrase-based system .

word sense disambiguation ( wsd ) is the task of identifying the correct meaning of a word in context .

word sense disambiguation ( wsd ) is a key enabling-technology .

language modeling is trained using kenlm using 5-grams , with modified kneser-ney smoothing .

the language models are 4-grams with modified kneser-ney smoothing which have been trained with the srilm toolkit .

the dataset we used in the present study is the online edition 2 of the world atlas of language structures .

we use the world atlas of language structure dataset , on which we conduct experiments .

modi et al extended the model of , which is an unsupervised model for inducing semantic roles , to jointly induce semantic roles and frames across verbs using the chinese restaurant process .

modi et al extended the model of to jointly induce semantic roles and frames using the chinese restaurant process , which is also used in our approach .

information extraction ( ie ) is the task of identifying information in texts and converting it into a predefined format .

information extraction ( ie ) is a technology that can be applied to identifying both sources and targets of new hyperlinks .

coreference resolution is a fundamental component of natural language processing ( nlp ) and has been widely applied in other nlp tasks ( cite-p-15-3-9 ) .

since coreference resolution is a pervasive discourse phenomenon causing performance impediments in current ie systems , we considered a corpus of aligned english and romanian texts to identify coreferring expressions .

for the lda based method , adding other content words , combined with an increased number of topics , can further improve the performance , achieving up to 14 . 23 % perplexity reduction .

in addition , we find that for the lda based adaptation scheme , adding more content words and increasing the number of topics can further improve the performance significantly .

sentiment analysis in twitter , which is a task of semeval , was firstly proposed in 2013 and not replaced until 2018 .

sentiment analysis in twitter is a particularly challenging task , because of the informal and “ creative ” writing style , with improper use of grammar , figurative language , misspellings and slang .

output string is guaranteed to conform to a given target grammar .

output always conforms to the given target grammar .

in this paper , we exploit non-local features as an estimate of long-distance dependencies .

in this paper , our approach describes how to exploit non-local information to a slu problem .

the smt systems were trained using the moses toolkit and the experiment management system .

the experiment management system from the open source moses smt toolkit was used to conduct the experiments .

these connections may be derived from work in language assessment and grade expectations such as found in , and .

these connections may be derived from work in language assessment and grade expectations such as found in .

the probabilistic language model is constructed on google web 1t 5-gram corpus by using the srilm toolkit .

a standard sri 5-gram language model is estimated from monolingual data .

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 ) .

word alignment is a critical first step for building statistical machine translation systems .

transition-based methods have given competitive accuracies and efficiencies for dependency parsing .

transition-based and graph-based models have attracted the most attention of dependency parsing in recent years .

zhuang et al present an algorithm for the extraction of opinion target -opinion word pairs .

zhuang et al present a supervised algorithm for the extraction of opinion expression -opinion target pairs .

therefore , we adopt the greedy feature selection algorithm as described in jiang et al to pick up positive features incrementally according to their contributions .

we adopt the greedy feature selection algorithm as described in jiang and ng to pick up positive features empirically and incrementally according to their contributions on the development data .

we estimated 5-gram language models using the sri toolkit with modified kneser-ney smoothing .

further , we apply a 4-gram language model trained with the srilm toolkit on the target side of the training corpus .

gold has been created to facilitate a more standardised use of basic grammatical features .

the idea behind gold is to facilitate a more standardised use of basic grammatical features .

therefore , we use the long short-term memory network to overcome this problem .

an effective solution for these problems is the long short-term memory architecture .

as described in this paper , we demonstrate the contribution of modality analysis for disease .

in this study , we use generic modality features to improve factuality analysis .

the first dataset is mov , which is a classical movie review dataset .

the first dataset is mov , which is a widely-used movie review dataset .

crfs have been shown to perform well on a number of nlp problems such as shallow parsing , table extraction , and named entity recognition .

crfs have been shown to perform well in a number of natural language processing applications , such as pos tagging , shallow parsing or np chunking , and named entity recognition .

ji and grishman employ an approach to propagate consistent event arguments across sentences and documents .

later , ji and grishman employed a rule-based approach to propagate consistent triggers and arguments across topic-related documents .

for training our system classifier , we have used scikit-learn .

we implement classification models using keras and scikit-learn .

the language models were built using srilm toolkits .

the srilm toolkit was used to build the 5-gram language model .

these embeddings are determined beforehand on a very large corpus typically using either the skip gram or the continuous bag of words variant of the word2vec model .

these word embeddings are learned in advance using a continuous skip-gram model , or other continuous word representation learning methods .

quirk et al extended path to treelets and put forward dependency treelet translation .

quirk et al and xiong et al used treelets to model the source dependency tree using synchronous grammars .

the spelling correction models from brill and moore and toutanova and moore use the noisy channel model approach to determine the types and weights of edit operations .

the spelling error model proposed by brill and moore allows generic string edit operations up to a certain length .

we use the sri language model toolkit to train a 5-gram model with modified kneser-ney smoothing on the target-side training corpus .

gram language model with modified kneser-ney smoothing is trained with the srilm toolkit on the epps , ted , newscommentary , and the gigaword corpora .

this is the first work that applies a state-of-the-art probabilistic parsing model to al for dependency parsing .

this paper for the first time applies a state-of-the-art probabilistic model to al with pa for dependency parsing .

the methods employed for gathering the data , preparation and compilation of dataset , used in offenseval shared task is described in zampieri et al .

the data collection methods used to compile the dataset used in offenseval is described in zampieri et al .

for evaluation we use mteval-v13a from the moses toolkit and tercom 3 to score our systems on the bleu respectively ter measures .

we use mteval from the moses toolkit an tercom to evaluate our systems on the bleu and ter measures .

bisk and hockenmaier used combinatory categorial grammar to learn syntactic dependencies from word strings .

bisk and hockenmaier use an em approach to induce a combinatory categorial grammar , based on very general linguistic assumptions .

the sg model is a popular choice to learn word embeddings by leveraging the relations between a word and its neighboring words .

the skip-gram model is a very popular technique for learning embeddings that scales to huge corpora and can capture important semantic and syntactic properties of words .

for word embeddings , we used popular pre-trained word vectors from glove .

we use the pre-trained glove vectors to initialize word embeddings .

wasserstein distance takes into account the cross-term relationship between different words in a principled fashion .

by using the wasserstein distance between distributions , the wordto-word semantic relationship is taken into account in a principled way .

eurowordnet is a multilingual lexical knowledge base comprised of hierarchical representations of lexical items for several european languages .

eurowordnet is a multilingual semantic lexicon with wordnets for several european languages , which are structured as the princeton wordnet .

klein and manning demonstrated that linguistically informed splitting of nonterminal symbols in treebank-derived grammars can result in accurate grammars .

klein and manning identified nonterminals that could valuably be split into fine-grained ones using hand-written linguistic rules .

sentiment classification is a well-studied and active research area ( cite-p-20-1-11 ) .

sentiment classification is a special task of text categorization that aims to classify documents according to their opinion of , or sentiment toward a given subject ( e.g. , if an opinion is supported or not ) ( cite-p-11-1-2 ) .

we will evaluate the performance of sampling distributions based on perplexities calculated using small , lightweight rnn language models .

we experimentally evaluate the heldout perplexity of models trained with our various importance sampling distributions .

for english , number and gender for common nouns are computed via a comparison of head lemma to head and using the number and gender data of bergsma and lin .

we compute number and gender for common nouns using the number and gender data provided by bergsma and lin .

a tree domain is a subset of strings over a linearly ordered set which is closed under prefix and left sister .

a tree domain is a set of node address drawn from n * ( that is , a set of strings of natural numbers ) in which c is the address of the root and the children of a node at address w occur at addresses w0 , wl , ... , in leftto-right order .

for our purpose we use word2vec embeddings trained on a google news dataset and find the pairwise cosine distances for all words .

we use the word2vec vectors with 300 dimensions , pre-trained on 100 billion words of google news .

ritchie et al used a combination of terms from citation contexts and existing index terms of a paper to improve indexing of cited papers .

for example , ritchie et al used a combination of terms from citation contexts and existing index terms of a paper to improve indexing of cited papers .

tag is a tree-rewriting system : the derivation process consists in applying operations to trees in order to obtain a ( derived ) tree whose sequence of leaves is a sentence .

a tag is a rewriting system that derives trees starting from a finite set of elementary trees .

we used implementations from scikitlearn , and the parameters of both classifiers were tuned on the development set using grid search .

we trained the three classifiers using the svm implementation in scikit-learn , and tuned hyper-parameters c and 纬 using 10-fold cross-validation with the train split .

dependency parsing is a simpler task than constituent parsing , since dependency trees do not have extra non-terminal nodes and there is no need for a grammar to generate them .

dependency parsing is the task of predicting the most probable dependency structure for a given sentence .

table 4 shows the bleu scores of the output descriptions .

the numbers in the table are bleu scores of different neural models .

we created 5-gram language models for every domain using srilm with improved kneserney smoothing on the target side of the training parallel corpora .

we train a 4-gram language model on the xinhua portion of the english gigaword corpus using the srilm toolkits with modified kneser-ney smoothing .

for parsing arabic texts into syntactic trees , we used the berkeley parser .

for our investigations , we used the berkeley parser as a source of grammar rule clusters .

in doing so we can achieve better word retrieval performance than language models with only n-gram context .

we aim to improve speech retrieval performance by augmenting traditional n-gram language models with different types of topic context .

svmhmms and crfs have been successfully applied to a range of sequential tagging tasks such as syllabification , chunk parsing and word segmentation .

crfs have been shown to perform well in a number of natural language processing applications , such as pos tagging , shallow parsing or np chunking , and named entity recognition .

in this paper , we name the problem of choosing the correct word from the homophone set .

in this paper , we used the decision list to solve the homophone problem .

we use the max-loss variant of the margin infused relaxed algorithm with a hamming-loss margin as is common in the dependency parsing literature .

we select the cutting-plane variant of the margin-infused relaxed algorithm with additional extensions described by eidelman .

we use a baseline parser to parse large-scale unannotated data .

first , we use a baseline parser to parse large-scale unannotated data .

system tuning was carried out using both k-best mira and minimum error rate training on the held-out development set .

feature weights were set with minimum error rate training on a development set using bleu as the objective function .

with the rise of social media , more and more user generated sentiment data have been shared on the web .

the rise of social media such as blogs and microblogs has fueled interest in sentiment analysis .

as a classifier , we choose a first-order conditional random field model .

to this end , we use first-and second-order conditional random fields .

mauser et al extended this model to condition it on source word cooccurrences .

mauser et al integrated a logistic regression model predicting target words from all the source words in a pbsmt .