HHousen/ParaSCI · Datasets at Hugging Face

sentence1 stringlengths 16 446	sentence2 stringlengths 14 436
in this paper , we propose a sense-topic model for wsi , which treats sense and topic as two separate latent variables to be inferred jointly .	we propose a sense-topic model for wsi , which treats sense and topic as two separate latent variables to be inferred jointly .
word space models capture the semantic similarity between two words on the basis of their distribution in a corpus .	co-occurrence space models represent the meaning of a word as a vector in high-dimensional space .
morante et al and morante and daelemans pioneered the research on negation scope finding by formulating it as a chunking problem , which classifies the words of a sentence as being inside or outside the scope of a negation signal .	morante et al and daelemans pioneered the research on scope learning by formulating it as a chunking problem , which classifies the words of a sentence as being inside or outside the scope of a cue .
trigram language models are implemented using the srilm toolkit .	the trigram language model is implemented in the srilm toolkit .
these models can be tuned using minimum error rate training .	each translation model is tuned using mert to maximize bleu .
language is the primary tool that people use for establishing , maintaining and expressing social relations .	language consists of much more than just content .
on the other hand , experiments indicate that mental representation and processing of morphologically complex words are not quite language independent .	on the other hand , several cross-linguistic experiments have indicated that mental representation and processing of polymorphemic words are not language independent .
the task of adapting a system trained on one domain ( called the source domain ) to a new domain ( called the target domain ) is called domain adaptation .	if these data sources differ systematically from each other , and/or from the test data , then the problem of combining these disparate data sets to create the best possible translation system is known as domain adaptation .
have shown that dual decomposition or lagrangian relaxation is an elegant framework for combining different types of nlp tasks .	it has recently been shown that different nlp models can be effectively combined using dual decomposition .
to the best of our knowledge , a large-scale quantitative typological analysis of lexical semantics is lacking thus far .	to our knowledge , this constitutes the first large-scale quantitative lexical semantic typology that is completely unsupervised , bottom-up , and data-driven .
relation detection is significantly more challenging compared to general relation detection tasks .	kb relation detection is a key step in kbqa and is significantly different from general relation extraction tasks .
jindal and liu base the recognition of comparative predicates on a list of manually compiled keywords .	jindal and liu introduce techniques to find and analyze explicit comparison sentences , but this assumes that such sentences exist .
one important difference between mstparser and maltparser , on the one hand , and the best performing parsers evaluated in rimell et al , on the other , is that the former were never developed specifically as parsers for english .	unlike the best performing grammar-based parsers studied in rimell et al , neither mstparser nor maltparser was developed specifically as a parser for english , and neither has any special mechanism for dealing with unbounded dependencies .
we used a phrase-based smt model as implemented in the moses toolkit .	we experimented with the phrase-based smt model as implemented in moses .
to generate the n-gram language models , we used the kenlm n-gram , language modeling tool .	we trained a standard 5-gram language model with modified kneser-ney smoothing using the kenlm toolkit on 4 billion running words .
the pre-processed monolingual sentences will be used by srilm or berkeleylm to train a n-gram language model .	a 5-gram language model with kneser-ney smoothing was trained with srilm on monolingual english data .
cite-p-17-3-2 proposed a recursive neural network designed to model the subtrees , and cnn to capture .	cite-p-17-1-2 proposed a simple customizaition of recursive neural networks .
sapkota et al showed that classical character n-grams lose some information in merging instances of ngrams like the which could be a prefix , a suffix , or a standalone word .	sapkota et al showed that classical character n-grams lose some information in merging together instances of n-grams like the which could be a prefix , a suffix , or a standalone word .
transitionbased and graph-based models have attracted the most attention of dependency parsing in recent years .	such approaches , for example , transition-based and graph-based models have attracted the most attention in dependency parsing in recent works .
this paper describes our investigation into the effectiveness of lexicalization in dependency parsing .	this paper discusses our investigation into the effectiveness of lexicalization in dependency parsing .
all the weights of those features are tuned by using minimal error rate training .	the weights associated to feature functions are optimally combined using the 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 use srilm to train a 5-gram language model on the target side of our training corpus with modified kneser-ney discounting .
for this language model , we built a trigram language model with kneser-ney smoothing using srilm from the same automatically segmented corpus .	for the language model we use the corpus of 60,000 simple english wikipedia articles 3 and build a 3-gram language model with kneser-ney smoothing trained with srilm .
in order to provide results on additional languages , we present in table 3 a comparison to the work of gillenwater et al , using the conll-x shared task data .	following pitler et al , we report in table 1 figures for the training sets of six languages used in the conll-x shared task on dependency parsing .
morphologically , arabic is a non-concatenative language .	arabic is a highly inflectional language with 85 % of words derived from trilateral roots ( alfedaghi and al-anzi 1989 ) .
language models were trained with the kenlm toolkit .	the 5-gram language models were built using kenlm .
twitter is the medium where people post real time messages to discuss on the different topics , and express their sentiments .	twitter is a very popular micro blogging site .
in this paper , we propose a feature augmentation approach for dependency parser adaptation .	in this paper , we present a parsing adaptation approach focused on the fully supervised case .
an english 5-gram language model is trained using kenlm on the gigaword corpus .	a 5-gram language model on the english side of the training data was trained with the kenlm toolkit .
time normalization is a crucial part of almost any information extraction task that needs to place entities or events along a timeline .	time normalization is the task of converting a natural language expression of time into a formal representation of a time on a timeline .
lda is a widely used topic model , which views the underlying document distribution as having a dirichlet prior .	lda is a probabilistic generative model that can be used to uncover the underlying semantic structure of a document collection .
word sense disambiguation ( wsd ) is a difficult natural language processing task which requires that for every content word ( noun , adjective , verb or adverb ) the appropriate meaning is automatically selected from the available sense inventory 1 .	word sense disambiguation ( wsd ) is the task to identify the intended sense of a word in a computational manner based on the context in which it appears ( cite-p-13-3-4 ) .
turney and littman proposed to compute pair-wised mutual information between a target word and a set of seed positive and negative words to infer the so of the target word .	turney and littman determined the polarity of sentiment words by estimating the point-wise mutual information between sentiment words and a set of seed words with strong polarity .
i like my x like i like my y , z jokes , where x , y , and z are variables to be filled in .	i like my x like i like my y , z , where x , y , and z are slots to be filled in .
the srilm toolkit was used to build this language model .	we used the sri language modeling toolkit for this purpose .
we used the sri language modeling toolkit to train a fivegram model with modified kneser-ney smoothing .	for the language model , we used sri language modeling toolkit to train a trigram model with modified kneser-ney smoothing on the 31 , 149 english sentences .
for phrase-based smt translation , we used the moses decoder and its support training scripts .	we used moses with the default configuration for phrase-based translation .
for estimating monolingual word vector models , we use the cbow algorithm as implemented in the word2vec package using a 5-token window .	to obtain these features , we use the word2vec implementation available in the gensim toolkit to obtain word vectors with dimension 300 for each word in the responses .
we use the moses software package 5 to train a pbmt model .	we implement the pbsmt system with the moses toolkit .
we used the svm-light-tk 5 to train the reranker with a combination of tree kernels and feature vectors .	we use svm-light-tk to train our reranking models , 9 which enables the use of tree kernels in svm-light .
silberer and frank use an entity-based coreference resolution model to automatically extended the training set .	silberer and frank cast ni resolution as a coreference resolution task , and employ an entity-mention model .
furthermore , we also evaluated the system on a similarly drug-focused corpus annotated for anaphora .	additionally , we tested our system on drugnerar corpus , which similarly focuses on drug interactions .
for this purpose , we use the uplug toolkit which is a collection of tools for processing corpus data , created by j枚rg tiedemann .	we use the uplug 5 collection of tools for alignment to extract translations from our specialized parallel corpus .
faruqui and dyer introduce canonical correlation analysis to project the embeddings in both languages to a shared vector space .	faruqui and dyer introduced canonical correlation analysis to project the embeddings in both languages to a shared vector space .
all smt models were developed using the moses phrase-based mt toolkit and the experiment management system .	both systems are phrase-based smt models , trained using the moses toolkit .
srilm toolkit is used to build these language models .	the trigram language model is implemented in the srilm toolkit .
relation extraction ( re ) is the task of determining semantic relations between entities mentioned in text .	relation extraction ( re ) is the task of extracting instances of semantic relations between entities in unstructured data such as natural language text .
we train trigram language models on the training set using the sri language modeling tookit .	we use sri language model toolkit to train a 5-gram model with modified kneser-ney smoothing on the target-side training corpus .
semantic parsing is the task of mapping a natural language ( nl ) sentence into a complete , formal meaning representation ( mr ) which a computer program can execute to perform some task , like answering database queries or controlling a robot .	semantic parsing is the problem of translating human language into computer language , and therefore is at the heart of natural language understanding .
system tuning was carried out using minimum error rate training optimised with k-best mira on a held out development set .	feature weights were set with minimum error rate training on a tuning set using bleu as the objective function .
word sense disambiguation ( wsd ) is a natural language processing ( nlp ) task in which the correct meaning ( sense ) of a word in a given context is to be determined .	word sense disambiguation ( wsd ) is a key enabling-technology that automatically chooses the intended sense of a word in context .
a hierarchical phrase-based translation grammar was extracted for the nist mt03 chinese-english translation using a suffix array rule extractor .	the hierarchical translation grammar was extracted using the joshua toolkit implementation of the suffix array rule extractor algorithm .
mutalik et al developed another rule based system called negfinder that recognizes negation patterns in biomedical text .	mutalik et al developed negfinder , a rule-based system that recognises negated patterns in medical documents .
in this paper , we adopt continuous bag-of-word in word2vec as our context-based embedding model .	with english gigaword corpus , we use the skip-gram model as implemented in word2vec 3 to induce embeddings .
cardie and wagstaff have proposed an unsupervised approach which also incorporates cluster information into consideration .	cardie and wagstaff present an early approach to unsupervised coreference resolution based on a straightforward clustering approach .
we use the skipgram model to learn word embeddings .	we use the monolingual corpora provided for the wmt translation task .
for our chinese-english experiments , we use a simple heuristic that equates anchors with constituents whose corresponding word class belongs to function words-related classes , bearing a close resemblance to .	for our chinese-english experiments , we use a simple heuristic that equates as anchors , single-word chunks whose corresponding word class belongs to closed-word classes , bearing a close resemblance to .
note that we interpret factuality as event factuality in the sense of saur铆 and pustejovsky .	following boye and saur铆 and pustejovsky , we characterize evidential justification in terms of epistemic support .
to test this hypothesis , we use a latent dirichlet allocation model .	for this feature , we use the latent dirichlet allocation .
these methods were normally created based on a large corpus of well-formed native english texts .	these methods are normally created based on a large corpus of well-formed native english texts .
we used moses as the implementation of the baseline smt systems .	we use the moses toolkit to train our phrase-based smt models .
system 1 is a new approach using sequence-tosequence models , encoderdecoder , and attention as described in bahdanau et al for machine translation .	our system for this shared task 1 is based on an encoder-decoder model proposed by bahdanau et al for neural machine translation .
conversational systems must be able to learn new words automatically during human machine conversation .	it is desirable that conversational systems can learn new words automatically during human machine conversation .
named entity recognition ( ner ) is the task of identifying and typing phrases that contain the names of persons , organizations , locations , and so on .	named entity recognition ( ner ) is a key technique for ie and other natural language processing tasks .
one way to handle such faulty arguments is to simply disregard them and focus on extracting arguments containing proper support .	one way to deal with such implicit arguments is to simply disregard them and focus on extracting arguments containing proper support .
ruppenhofer et al argued that semantic role techniques are useful but not completely sufficient for holder and topic identification , and that other linguistic phenomena must be studied as well .	ruppenhofer , somasundaran , and wiebe argued that semantic role techniques are useful but not completely sufficient for holder and topic identification , and that other linguistic phenomena must be studied as well .
recurrent neural networks are another way to exploit the context of a word by considering the sequence of words preceding it .	recurrent neural networks are another natural choice to model text due to their capability of processing arbitrary-length sequences .
this paper presents a thorough evaluation of the impact of annotation noise on al .	in this paper we present a thorough evaluation of the impact of annotation noise on al .
relation classification is the task of finding semantic relations between pairs of nominals , which is useful for many nlp applications , such as information extraction ( cite-p-15-3-3 ) , question answering ( cite-p-15-3-6 ) .	relation classification is the task of identifying the semantic relation holding between two nominal entities in text .
then , additional alignment points are added according to the growing heuristic algorithm , grow additional alignment points , finally , we select consecutive clusters which are aligned to the same english word as candidates .	then , additional alignment points are added according to the growing heuristic algorithm , grow additional alignment points , finally , we select consecutive which are aligned to the same english word as candidates .
liu et al and baron et al carried out sentence unit and disfluency prediction as separate tasks .	liu et al used conditional random fields for sentence boundary and edit word detection .
in cross-lingual settings , the actual translations of a word can be taken as the sense labels .	in a multilingual context , word senses can be easily identified using their translations in other languages .
we train a linear support vector machine classifier using the efficient liblinear package .	we build all the classifiers using the l2-regularized linear logistic regression from the liblinear package .
in this study , we presented a novel encoder-decoder model to automatically generate market comments from numerical time-series data of stock prices .	in this study , we focus on the task of generating market comments from a time-series of stock prices .
semantic similarity has seen major progress in recent times , due largely to the semeval semantic textual similarity ( sts ) task ( cite-p-17-1-0 , cite-p-17-1-1 , cite-p-17-1-2 , cite-p-17-1-3 ) .	major progress has been made in this task in recent years , due primarily to the semeval semantic textual similarity ( sts ) task ( cite-p-17-1-0 , cite-p-17-1-1 , cite-p-17-1-2 , cite-p-17-1-3 ) .
hamilton et al propose the use of cosine similarities of words in different contexts to detect changes .	hamilton et al measured the variation between models by observing semantic change using diachronic corpora .
unfortunately , wordnet is a fine-grained resource , which encodes possibly subtle sense distictions .	wordnet is a comprehensive lexical resource for word-sense disambiguation ( wsd ) , covering nouns , verbs , adjectives , adverbs , and many multi-word expressions .
the dataset we used in the present study is the online edition 2 of the world atlas of language structures .	the database of typological features we used is the online edition 8 of the world atlas of language structures .
experimental results demonstrate that the distributional similarity based models can significantly outperform their baseline systems .	experimental results show that both methods can achieve significant improvements over their baseline settings .
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 .	language models were estimated using the sri language modeling toolkit with modified kneser-ney smoothing .
translation quality is measured in truecase with bleu and ter .	the translation quality is evaluated by case-insensitive bleu and ter metrics using multeval .
also , convolutional neural networks have been a popular choice in the image domain .	convolutional neural networks have rapidly become the state-of-the-art approach in computer vision .
we employ support vector machine as the machine learning approach .	in the experiments reported here we use support vector machines through the svm light package .
experiments on chinese-english translation show that joint training with generalized agreement achieves significant improvements over two baselines for ( hierarchical ) .	experiments on chineseenglish translation show that joint training with generalized agreement achieves significant improvements over two state-of-the-art alignment methods .
infrastructural issues are dealt with by the platform , completely transparently to the user : load balancing , efficient data upload and storage , deployment on the virtual machines , security , and fault tolerance .	important infrastructural issues are dealt with by the platform , completely transparently for the researcher : load balancing , efficient data upload and storage , deployment on the virtual machines , security , and fault tolerance .
hence we use the expectation maximization algorithm for parameter learning .	a minimum of this function can be found using the em algorithm .
for language model , we used sri language modeling toolkit to train a 4-gram model with modified kneser-ney smoothing .	we also use a 4-gram language model trained using srilm with kneser-ney smoothing .
we develop translation models using the phrase-based moses smt system .	our baseline system is phrase-based moses with feature weights trained using mert .
figure 1 : multimodal compact bilinear pooling .	figure 2 : multimodal compact bilinear pooling ( mcb )
madamira is a tool designed for morphological analysis and disambiguation of modern standard arabic .	madamira is a system for morphological analysis and disambiguation of arabic text .
goldwasser et al presented a confidence-driven approach to semantic parsing based on self-training .	goldwasser et al took an unsupervised approach for semantic parsing based on self-training driven by confidence estimation .
this paper proposes the ¡° hierarchical directed acyclic graph ( hdag ) kernel .	this paper proposes the hierarchical directed acyclic graph ( hdag ) kernel .
also , we initialized all of the word embeddings using the 300 dimensional pre-trained vectors from glove .	our word embeddings is initialized with 100-dimensional glove word embeddings .
we also use glove vectors to initialize the word embedding matrix in the caption embedding module .	we use pre-trained glove vector for initialization of word embeddings .
for instance , collobert and weston use a multitask network for different nlp tasks and show that the multi-task setting improves generality among shared tasks .	collobert and weston propose a unified deep convolutional neural network for different tasks by using a set of taskindependent word embeddings together with a set of task-specific word embeddings .
with dependency treebanks , since ccg lexical categories can be easily extracted from dependency treebanks ( cite-p-19-1-2 , cite-p-19-1-0 ) .	our approach can therefore be adapted to languages with dependency treebanks , since ccg lexical categories can be easily extracted from dependency treebanks ( cite-p-19-1-2 , cite-p-19-1-0 ) .
in this paper , we present a reranking method of n-best multi-sentence compressions based on keyphrase extraction .	this paper presents an n-best reranking method based on keyphrase extraction .
the default is the phrase-based variant of cube pruning .	the default phrasal search algorithm is cube pruning .
axelrod et al , 2011 ) extend this work using the bilingual difference of cross-entropy on in-domain and out-of-domain language models for training sentence selection for smt .	axelrod et al improved the perplexitybased approach and proposed bilingual crossentropy difference as a ranking function with inand general-domain language models .
semantic role labeling ( srl ) is the task of automatic recognition of individual predicates together with their major roles ( e.g . frame elements ) as they are grammatically realized in input sentences .	semantic role labeling ( srl ) is the task of identifying the semantic arguments of a predicate and labeling them with their semantic roles .

in this paper , we propose a sense-topic model for wsi , which treats sense and topic as two separate latent variables to be inferred jointly .

we propose a sense-topic model for wsi , which treats sense and topic as two separate latent variables to be inferred jointly .

word space models capture the semantic similarity between two words on the basis of their distribution in a corpus .

co-occurrence space models represent the meaning of a word as a vector in high-dimensional space .

morante et al and morante and daelemans pioneered the research on negation scope finding by formulating it as a chunking problem , which classifies the words of a sentence as being inside or outside the scope of a negation signal .

morante et al and daelemans pioneered the research on scope learning by formulating it as a chunking problem , which classifies the words of a sentence as being inside or outside the scope of a cue .

trigram language models are implemented using the srilm toolkit .

the trigram language model is implemented in the srilm toolkit .

these models can be tuned using minimum error rate training .

each translation model is tuned using mert to maximize bleu .

language is the primary tool that people use for establishing , maintaining and expressing social relations .

language consists of much more than just content .

on the other hand , experiments indicate that mental representation and processing of morphologically complex words are not quite language independent .

on the other hand , several cross-linguistic experiments have indicated that mental representation and processing of polymorphemic words are not language independent .

the task of adapting a system trained on one domain ( called the source domain ) to a new domain ( called the target domain ) is called domain adaptation .

if these data sources differ systematically from each other , and/or from the test data , then the problem of combining these disparate data sets to create the best possible translation system is known as domain adaptation .

have shown that dual decomposition or lagrangian relaxation is an elegant framework for combining different types of nlp tasks .

it has recently been shown that different nlp models can be effectively combined using dual decomposition .

to the best of our knowledge , a large-scale quantitative typological analysis of lexical semantics is lacking thus far .

to our knowledge , this constitutes the first large-scale quantitative lexical semantic typology that is completely unsupervised , bottom-up , and data-driven .

relation detection is significantly more challenging compared to general relation detection tasks .

kb relation detection is a key step in kbqa and is significantly different from general relation extraction tasks .

jindal and liu base the recognition of comparative predicates on a list of manually compiled keywords .

jindal and liu introduce techniques to find and analyze explicit comparison sentences , but this assumes that such sentences exist .

one important difference between mstparser and maltparser , on the one hand , and the best performing parsers evaluated in rimell et al , on the other , is that the former were never developed specifically as parsers for english .

unlike the best performing grammar-based parsers studied in rimell et al , neither mstparser nor maltparser was developed specifically as a parser for english , and neither has any special mechanism for dealing with unbounded dependencies .

we used a phrase-based smt model as implemented in the moses toolkit .

we experimented with the phrase-based smt model as implemented in moses .

to generate the n-gram language models , we used the kenlm n-gram , language modeling tool .

we trained a standard 5-gram language model with modified kneser-ney smoothing using the kenlm toolkit on 4 billion running words .

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

a 5-gram language model with kneser-ney smoothing was trained with srilm on monolingual english data .

cite-p-17-3-2 proposed a recursive neural network designed to model the subtrees , and cnn to capture .

cite-p-17-1-2 proposed a simple customizaition of recursive neural networks .

sapkota et al showed that classical character n-grams lose some information in merging instances of ngrams like the which could be a prefix , a suffix , or a standalone word .

sapkota et al showed that classical character n-grams lose some information in merging together instances of n-grams like the which could be a prefix , a suffix , or a standalone word .

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

such approaches , for example , transition-based and graph-based models have attracted the most attention in dependency parsing in recent works .

this paper describes our investigation into the effectiveness of lexicalization in dependency parsing .

this paper discusses our investigation into the effectiveness of lexicalization in dependency parsing .

all the weights of those features are tuned by using minimal error rate training .

the weights associated to feature functions are optimally combined using the 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 use srilm to train a 5-gram language model on the target side of our training corpus with modified kneser-ney discounting .

for this language model , we built a trigram language model with kneser-ney smoothing using srilm from the same automatically segmented corpus .

for the language model we use the corpus of 60,000 simple english wikipedia articles 3 and build a 3-gram language model with kneser-ney smoothing trained with srilm .

in order to provide results on additional languages , we present in table 3 a comparison to the work of gillenwater et al , using the conll-x shared task data .

following pitler et al , we report in table 1 figures for the training sets of six languages used in the conll-x shared task on dependency parsing .

morphologically , arabic is a non-concatenative language .

arabic is a highly inflectional language with 85 % of words derived from trilateral roots ( alfedaghi and al-anzi 1989 ) .

language models were trained with the kenlm toolkit .

the 5-gram language models were built using kenlm .

twitter is the medium where people post real time messages to discuss on the different topics , and express their sentiments .

twitter is a very popular micro blogging site .

in this paper , we propose a feature augmentation approach for dependency parser adaptation .

in this paper , we present a parsing adaptation approach focused on the fully supervised case .

an english 5-gram language model is trained using kenlm on the gigaword corpus .

a 5-gram language model on the english side of the training data was trained with the kenlm toolkit .

time normalization is a crucial part of almost any information extraction task that needs to place entities or events along a timeline .

time normalization is the task of converting a natural language expression of time into a formal representation of a time on a timeline .

lda is a widely used topic model , which views the underlying document distribution as having a dirichlet prior .

lda is a probabilistic generative model that can be used to uncover the underlying semantic structure of a document collection .

word sense disambiguation ( wsd ) is a difficult natural language processing task which requires that for every content word ( noun , adjective , verb or adverb ) the appropriate meaning is automatically selected from the available sense inventory 1 .

word sense disambiguation ( wsd ) is the task to identify the intended sense of a word in a computational manner based on the context in which it appears ( cite-p-13-3-4 ) .

turney and littman proposed to compute pair-wised mutual information between a target word and a set of seed positive and negative words to infer the so of the target word .

turney and littman determined the polarity of sentiment words by estimating the point-wise mutual information between sentiment words and a set of seed words with strong polarity .

i like my x like i like my y , z jokes , where x , y , and z are variables to be filled in .

i like my x like i like my y , z , where x , y , and z are slots to be filled in .

the srilm toolkit was used to build this language model .

we used the sri language modeling toolkit for this purpose .

we used the sri language modeling toolkit to train a fivegram model with modified kneser-ney smoothing .

for the language model , we used sri language modeling toolkit to train a trigram model with modified kneser-ney smoothing on the 31 , 149 english sentences .

for phrase-based smt translation , we used the moses decoder and its support training scripts .

we used moses with the default configuration for phrase-based translation .

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

to obtain these features , we use the word2vec implementation available in the gensim toolkit to obtain word vectors with dimension 300 for each word in the responses .

we use the moses software package 5 to train a pbmt model .

we implement the pbsmt system with the moses toolkit .

we used the svm-light-tk 5 to train the reranker with a combination of tree kernels and feature vectors .

we use svm-light-tk to train our reranking models , 9 which enables the use of tree kernels in svm-light .

silberer and frank use an entity-based coreference resolution model to automatically extended the training set .

silberer and frank cast ni resolution as a coreference resolution task , and employ an entity-mention model .

furthermore , we also evaluated the system on a similarly drug-focused corpus annotated for anaphora .

additionally , we tested our system on drugnerar corpus , which similarly focuses on drug interactions .

for this purpose , we use the uplug toolkit which is a collection of tools for processing corpus data , created by j枚rg tiedemann .

we use the uplug 5 collection of tools for alignment to extract translations from our specialized parallel corpus .

faruqui and dyer introduce canonical correlation analysis to project the embeddings in both languages to a shared vector space .

faruqui and dyer introduced canonical correlation analysis to project the embeddings in both languages to a shared vector space .

all smt models were developed using the moses phrase-based mt toolkit and the experiment management system .

both systems are phrase-based smt models , trained using the moses toolkit .

srilm toolkit is used to build these language models .

the trigram language model is implemented in the srilm toolkit .

relation extraction ( re ) is the task of determining semantic relations between entities mentioned in text .

relation extraction ( re ) is the task of extracting instances of semantic relations between entities in unstructured data such as natural language text .

we train trigram language models on the training set using the sri language modeling tookit .

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

semantic parsing is the task of mapping a natural language ( nl ) sentence into a complete , formal meaning representation ( mr ) which a computer program can execute to perform some task , like answering database queries or controlling a robot .

semantic parsing is the problem of translating human language into computer language , and therefore is at the heart of natural language understanding .

system tuning was carried out using minimum error rate training optimised with k-best mira on a held out development set .

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

word sense disambiguation ( wsd ) is a natural language processing ( nlp ) task in which the correct meaning ( sense ) of a word in a given context is to be determined .

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

a hierarchical phrase-based translation grammar was extracted for the nist mt03 chinese-english translation using a suffix array rule extractor .

the hierarchical translation grammar was extracted using the joshua toolkit implementation of the suffix array rule extractor algorithm .

mutalik et al developed another rule based system called negfinder that recognizes negation patterns in biomedical text .

mutalik et al developed negfinder , a rule-based system that recognises negated patterns in medical documents .

in this paper , we adopt continuous bag-of-word in word2vec as our context-based embedding model .

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

cardie and wagstaff have proposed an unsupervised approach which also incorporates cluster information into consideration .

cardie and wagstaff present an early approach to unsupervised coreference resolution based on a straightforward clustering approach .

we use the skipgram model to learn word embeddings .

we use the monolingual corpora provided for the wmt translation task .

for our chinese-english experiments , we use a simple heuristic that equates anchors with constituents whose corresponding word class belongs to function words-related classes , bearing a close resemblance to .

for our chinese-english experiments , we use a simple heuristic that equates as anchors , single-word chunks whose corresponding word class belongs to closed-word classes , bearing a close resemblance to .

note that we interpret factuality as event factuality in the sense of saur铆 and pustejovsky .

following boye and saur铆 and pustejovsky , we characterize evidential justification in terms of epistemic support .

to test this hypothesis , we use a latent dirichlet allocation model .

for this feature , we use the latent dirichlet allocation .

these methods were normally created based on a large corpus of well-formed native english texts .

these methods are normally created based on a large corpus of well-formed native english texts .

we used moses as the implementation of the baseline smt systems .

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

system 1 is a new approach using sequence-tosequence models , encoderdecoder , and attention as described in bahdanau et al for machine translation .

our system for this shared task 1 is based on an encoder-decoder model proposed by bahdanau et al for neural machine translation .

conversational systems must be able to learn new words automatically during human machine conversation .

it is desirable that conversational systems can learn new words automatically during human machine conversation .

named entity recognition ( ner ) is the task of identifying and typing phrases that contain the names of persons , organizations , locations , and so on .

named entity recognition ( ner ) is a key technique for ie and other natural language processing tasks .

one way to handle such faulty arguments is to simply disregard them and focus on extracting arguments containing proper support .

one way to deal with such implicit arguments is to simply disregard them and focus on extracting arguments containing proper support .

ruppenhofer et al argued that semantic role techniques are useful but not completely sufficient for holder and topic identification , and that other linguistic phenomena must be studied as well .

ruppenhofer , somasundaran , and wiebe argued that semantic role techniques are useful but not completely sufficient for holder and topic identification , and that other linguistic phenomena must be studied as well .

recurrent neural networks are another way to exploit the context of a word by considering the sequence of words preceding it .

recurrent neural networks are another natural choice to model text due to their capability of processing arbitrary-length sequences .

this paper presents a thorough evaluation of the impact of annotation noise on al .

in this paper we present a thorough evaluation of the impact of annotation noise on al .

relation classification is the task of finding semantic relations between pairs of nominals , which is useful for many nlp applications , such as information extraction ( cite-p-15-3-3 ) , question answering ( cite-p-15-3-6 ) .

relation classification is the task of identifying the semantic relation holding between two nominal entities in text .

then , additional alignment points are added according to the growing heuristic algorithm , grow additional alignment points , finally , we select consecutive clusters which are aligned to the same english word as candidates .

then , additional alignment points are added according to the growing heuristic algorithm , grow additional alignment points , finally , we select consecutive which are aligned to the same english word as candidates .

liu et al and baron et al carried out sentence unit and disfluency prediction as separate tasks .

liu et al used conditional random fields for sentence boundary and edit word detection .

in cross-lingual settings , the actual translations of a word can be taken as the sense labels .

in a multilingual context , word senses can be easily identified using their translations in other languages .

we train a linear support vector machine classifier using the efficient liblinear package .

we build all the classifiers using the l2-regularized linear logistic regression from the liblinear package .

in this study , we presented a novel encoder-decoder model to automatically generate market comments from numerical time-series data of stock prices .

in this study , we focus on the task of generating market comments from a time-series of stock prices .

semantic similarity has seen major progress in recent times , due largely to the semeval semantic textual similarity ( sts ) task ( cite-p-17-1-0 , cite-p-17-1-1 , cite-p-17-1-2 , cite-p-17-1-3 ) .

major progress has been made in this task in recent years , due primarily to the semeval semantic textual similarity ( sts ) task ( cite-p-17-1-0 , cite-p-17-1-1 , cite-p-17-1-2 , cite-p-17-1-3 ) .

hamilton et al propose the use of cosine similarities of words in different contexts to detect changes .

hamilton et al measured the variation between models by observing semantic change using diachronic corpora .

unfortunately , wordnet is a fine-grained resource , which encodes possibly subtle sense distictions .

wordnet is a comprehensive lexical resource for word-sense disambiguation ( wsd ) , covering nouns , verbs , adjectives , adverbs , and many multi-word expressions .

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

the database of typological features we used is the online edition 8 of the world atlas of language structures .

experimental results demonstrate that the distributional similarity based models can significantly outperform their baseline systems .

experimental results show that both methods can achieve significant improvements over their baseline settings .

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 .

language models were estimated using the sri language modeling toolkit with modified kneser-ney smoothing .

translation quality is measured in truecase with bleu and ter .

the translation quality is evaluated by case-insensitive bleu and ter metrics using multeval .

also , convolutional neural networks have been a popular choice in the image domain .

convolutional neural networks have rapidly become the state-of-the-art approach in computer vision .

we employ support vector machine as the machine learning approach .

in the experiments reported here we use support vector machines through the svm light package .

experiments on chinese-english translation show that joint training with generalized agreement achieves significant improvements over two baselines for ( hierarchical ) .

experiments on chineseenglish translation show that joint training with generalized agreement achieves significant improvements over two state-of-the-art alignment methods .

infrastructural issues are dealt with by the platform , completely transparently to the user : load balancing , efficient data upload and storage , deployment on the virtual machines , security , and fault tolerance .

important infrastructural issues are dealt with by the platform , completely transparently for the researcher : load balancing , efficient data upload and storage , deployment on the virtual machines , security , and fault tolerance .

hence we use the expectation maximization algorithm for parameter learning .

a minimum of this function can be found using the em algorithm .

for language model , we used sri language modeling toolkit to train a 4-gram model with modified kneser-ney smoothing .

we also use a 4-gram language model trained using srilm with kneser-ney smoothing .

we develop translation models using the phrase-based moses smt system .

our baseline system is phrase-based moses with feature weights trained using mert .

figure 1 : multimodal compact bilinear pooling .

figure 2 : multimodal compact bilinear pooling ( mcb )

madamira is a tool designed for morphological analysis and disambiguation of modern standard arabic .

madamira is a system for morphological analysis and disambiguation of arabic text .

goldwasser et al presented a confidence-driven approach to semantic parsing based on self-training .

goldwasser et al took an unsupervised approach for semantic parsing based on self-training driven by confidence estimation .

this paper proposes the ¡° hierarchical directed acyclic graph ( hdag ) kernel .

this paper proposes the hierarchical directed acyclic graph ( hdag ) kernel .

also , we initialized all of the word embeddings using the 300 dimensional pre-trained vectors from glove .

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

we also use glove vectors to initialize the word embedding matrix in the caption embedding module .

we use pre-trained glove vector for initialization of word embeddings .

for instance , collobert and weston use a multitask network for different nlp tasks and show that the multi-task setting improves generality among shared tasks .

collobert and weston propose a unified deep convolutional neural network for different tasks by using a set of taskindependent word embeddings together with a set of task-specific word embeddings .

with dependency treebanks , since ccg lexical categories can be easily extracted from dependency treebanks ( cite-p-19-1-2 , cite-p-19-1-0 ) .

our approach can therefore be adapted to languages with dependency treebanks , since ccg lexical categories can be easily extracted from dependency treebanks ( cite-p-19-1-2 , cite-p-19-1-0 ) .

in this paper , we present a reranking method of n-best multi-sentence compressions based on keyphrase extraction .

this paper presents an n-best reranking method based on keyphrase extraction .

the default is the phrase-based variant of cube pruning .

the default phrasal search algorithm is cube pruning .

axelrod et al , 2011 ) extend this work using the bilingual difference of cross-entropy on in-domain and out-of-domain language models for training sentence selection for smt .

axelrod et al improved the perplexitybased approach and proposed bilingual crossentropy difference as a ranking function with inand general-domain language models .

semantic role labeling ( srl ) is the task of automatic recognition of individual predicates together with their major roles ( e.g . frame elements ) as they are grammatically realized in input sentences .

semantic role labeling ( srl ) is the task of identifying the semantic arguments of a predicate and labeling them with their semantic roles .