sentence1
stringlengths
16
446
sentence2
stringlengths
14
436
we used translated movie subtitles from the freely available opus corpus .
we chose to use data from a collection of translated subtitles compiled in the freely available opus corpus .
zhao et al use a background topic in twitter-lda to distill discriminative words in tweets .
zhao et al used topic modelling to compare twitter with the new york times news site .
lam et al present work on email summarization by exploiting the thread structure of email conversation and common features such as named entities and dates .
lam et al , 2002 ) present work on email summarization by exploiting the thread structure of email conversation and common features such as named entities and dates .
we built a 5-gram language model on the english side of europarl and used the kneser-ney smoothing method and srilm as the language model toolkit .
we used europarl and wikipedia as parallel resources and all of the finnish data available from wmt to train five-gram language models with srilm and kenlm .
kilicoglu and bergler proposed a linguistically motivated approach based on syntactic information to semi-automatically refine a list of hedge cues .
kilicoglu and bergler apply a linguistically motivated approach to the same classification task by using knowledge from existing lexical resources and incorporating syntactic patterns .
amr is a semantic formalism , structured as a graph ( cite-p-13-1-1 ) .
amr is a graph representation for the meaning of a sentence , in which noun phrases ( nps ) are manually annotated with internal structure and semantic relations .
for building our statistical ape system , we used maximum phrase length of 7 and a 5-gram language model trained using kenlm .
in addition , we use an english corpus of roughly 227 million words to build a target-side 5-gram language model with srilm in combination with kenlm .
as such , masc is the first large-scale , open , community-based effort to create a much-needed language resource for nlp .
as such , masc is the first large-scale , open , community-based effort to create much needed language resources for nlp .
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 .
coreference resolution is the task of automatically grouping references to the same real-world entity in a document into a set .
we trained a 5-gram language model on the xinhua portion of gigaword corpus using the srilm toolkit .
we trained a 5-gram sri language model using the corpus supplied for this purpose by the shared task organizers .
generative models like lda and plsa have been proved to be very successful in modeling topics and other textual information in an unsupervised manner .
topic models , such as plsa and lda , have shown great success in discovering latent topics in text collections .
sentiment analysis is a technique to classify documents based on the polarity of opinion expressed by the author of the document ( cite-p-16-1-13 ) .
sentiment analysis is the study of the subjectivity and polarity ( positive vs. negative ) of a text ( cite-p-7-1-10 ) .
we use the pre-trained glove 50-dimensional word embeddings to represent words found in the glove dataset .
we use the glove pre-trained word embeddings for the vectors of the content words .
for instance , choudhury et al predicted the onset of depression from user tweets , while other studies have modeled distress .
for instance , de choudhury et al predicted the onset of depression from user tweets , while other studies have modeled distress .
luong et al segment words using morfessor , and use recursive neural networks to build word embeddings from morph embeddings .
luong et al utilized the morpheme segments produced by morfessor and constructed morpheme trees for words to learn morphologically-aware word embeddings by the recursive neural network .
for this language , which has limited the number of possible tags , we used a very rich tagset of 680 morphosyntactic tags .
unlike most previous work , which has used a reduced set of pos tags , we use all 680 tags in the bultreebank .
twitter is a well-known social network service that allows users to post short 140 character status update which is called “ tweet ” .
twitter is a microblogging service that has 313 million monthly active users 1 .
we apply the stochastic gradient descent algorithm with mini-batches and the adadelta update rule .
we train the model by using a simple optimization technique called stochastic gradient descent over shuffled mini-batches with the adadelta rule .
meta-analytic findings indicate that human judges are just slightly better than chance at discriminating between truths and lies .
human judges are often only slightly better than chance at discriminating between truths and lies .
semantic role labeling ( srl ) is the process of assigning semantic roles to strings of words in a sentence according to their relationship to the semantic predicates expressed in the sentence .
semantic role labeling ( srl ) has been defined as a sentence-level natural-language processing task in which semantic roles are assigned to the syntactic arguments of a predicate ( cite-p-14-1-7 ) .
in order to directly optimize the task reward of the structured prediction problem .
we apply reinforce to directly optimize the task reward of this structured prediction problem .
vocabulary lists were drawn from the french component of the unified medical language system and the vi-dal drug database .
items were identified using the unified medical language system vocabularies for dictionary look-up .
character-based and word-based ner methods , our model has the advantage of leveraging explicit word information over character sequence labeling .
compared with character-based methods , our model explicitly leverages word and word sequence information .
for training the translation model and for decoding we used the moses toolkit .
we use moses , a statistical machine translation system that allows training of translation models .
the model parameters of word embedding are initialized using word2vec .
the word embeddings are initialized with 100-dimensions vectors pre-trained by the cbow model .
to evaluate the k-qard framework , we built restricted domain question answering systems .
we present a korean question answering framework for restricted domains , called k-qard .
the encoder units are bidirectional lstms while the decoder unit incorporates an lstm with dot product attention .
the decoder and encoder word embeddings are of size 500 , the encoder uses a bidirectional lstm layer with 1k units to encode the source side .
we used a logistic regression classifier provided by the liblinear software .
we use liblinear logistic regression module to classify document-level embeddings .
in the introduction , these methods do not focus on the ability to accumulate learned knowledge and leverage it in new learning .
however , as we discussed in the introduction , these methods do not focus on the ability to accumulate learned knowledge and leverage it in new learning in a lifelong manner .
the language models are 4-grams with modified kneser-ney smoothing which have been trained with the srilm toolkit .
the language model was constructed using the srilm toolkit with interpolated kneser-ney discounting .
koppel et al did suggest using syntactic errors in their work but did not investigate them in any detail .
koppel et al also suggest that syntactic errors might be useful features , but these were not explored in their study .
the system automatically generates a thesaurus using a measure of distributional similarity and an untagged corpus .
to generate the greatest breadth of synonyms , the tool uses a distributional thesaurus , wordnet and a paraphrase generation tool .
coreference resolution is the process of finding discourse entities ( markables ) referring to the same real-world entity or concept .
coreference resolution is the process of linking together multiple expressions of a given entity .
berant et al proposed a semantic parsing model that can be trained from qna pairs , which are much easier to obtain than correct kb queries used previously .
berant et al have used a lexicon extracted from a subset of reverb triples , which is similar to the relation expression set used in question translation .
word sense disambiguation ( wsd ) is the task of assigning sense tags to ambiguous lexical items ( lis ) in a text .
word sense disambiguation ( wsd ) is the task of determining the correct meaning ( “ sense ” ) of a word in context , and several efforts have been made to develop automatic wsd systems .
we compare bi-lstms to traditional pos taggers .
we consider using bi-lstms for pos tagging .
the kit translations are generated by an in-house phrase-based translations system .
the in-house phrase-based translation system is used for generating translations .
we show that such a system provides an accuracy rivaling that of experts .
our approach shows an accuracy that rivals that of expert agreement .
auli et al presented a joint language and translation model based on a recurrent neural network which predicts target words based on an unbounded history of both source and target words .
auli et al and kalchbrenner and blunsom proposed joint language and translation model with recurrent neural networks , in which latent semantic analysis and convolutional sentence model were used to model source-side sentence .
sentiment analysis is a multi-faceted problem .
sentiment analysis is the study of the subjectivity and polarity ( positive vs. negative ) of a text ( cite-p-7-1-10 ) .
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 calculate the pointwise mutual information of a given word with positive and negative sets of sentiment words .
takamura et al proposed using spin models for extracting semantic orientation of words .
takamura et al propose using spin models for extracting semantic orientation of words .
parallel data in the domain of interest is the key resource when training a statistical machine translation ( smt ) system for a specific business purpose .
parallel data in the domain of interest is the key resource when training a statistical machine translation ( smt ) system for a specific purpose .
in our word embedding training , we use the word2vec implementation of skip-gram .
our cdsm feature is based on word vectors derived using a skip-gram model .
we report bleu scores to compare translation results .
we use bleu to evaluate translation quality .
we use the sri language model toolkit to train a 5-gram model with modified kneser-ney smoothing on the target-side training corpus .
we used kenlm with srilm to train a 5-gram language model based on all available target language training data .
as for the boundary detection problem , we use the windowdiff and p k metrics .
we classically used the error metric p k proposed in and its variant windowdiff to measure segmentation accuracy .
all models used interpolated modified kneser-ney smoothing .
the language model is a 5-gram with interpolation and kneser-ney smoothing .
but rather than solely making using of sparse binary features , we explicitly model dependency paths .
by modeling dependency paths as sequences of words and dependencies , we implicitly address the data sparsity problem .
peters et al show that their language model elmo can implicitly disambiguate word meaning with their contexts .
peters et al propose a deep neural model that generates contextual word embeddings which are able to model both language and semantics of word use .
the weights of the linear ranker are optimized using the averaged perceptron algorithm .
we train a linear classifier using the averaged perceptron algorithm .
we use manually and automatically determined sentiment labels of the arabic tweets .
as benchmarks , we use manually and automatically determined sentiment labels of the arabic texts .
in this paper we introduced sepa , a novel algorithm for assessing parse quality .
in this paper we present a sample ensemble parse assessment ( sepa ) algorithm for detecting parse quality .
word sense disambiguation ( wsd ) is the task of automatically determining the correct sense for a target word given the context in which it occurs .
word sense disambiguation ( wsd ) is the task of determining the meaning of an ambiguous word in its context .
we make use of an ensemble of text only attention based nmt models with a conditional gated recurrent units decoder .
we use the attention-based nmt model introduced by bahdanau et al as our text-only nmt baseline .
and we also illustrate how our model correctly identifies ideological bias .
in addition , we describe an approach to crowdsourcing ideological bias annotations .
user simulations are commonly used to train strategies for dialogue management , see for example .
user simulations are commonly used to train strategies for dialogue management , see for example .
the use of various synchronous grammar based formalisms has been a trend for statistical machine translation .
hierarchical phrase-based translation models that utilize synchronous context free grammars have been widely adopted in statistical machine translation .
we used the sri language modeling toolkit to train a fivegram model with modified kneser-ney smoothing .
we use srilm to train a 5-gram language model on the target side of our training corpus with modified kneser-ney discounting .
chen et al , 2012 ) proposed the lexical syntactic feature architecture to detect offensive content and identify the potential offensive users in social media .
chen et al introduced a lexical syntactic feature architecture to detect offensive content and identify potential offensive users in social media .
coreference resolution is a challenging task , that involves identification and clustering of noun phrases mentions that refer to the same real-world entity .
coreference resolution is the task of identifying all mentions which refer to the same entity in a document .
recently , distributional features have also been used directly to train classifiers that classify pairs of words as being synonymous or not .
in addition , recently , distributional features have also been used directly to train classifiers that classify pairs of words as being synonymous or not and showed good performance on the applied tasks .
a pseudoword is a composite comprised of two or more words chosen at random ; the individual occurrences of the original words within a text are replaced by their conflation .
pseudo-word is a kind of multi-word expression ( includes both unary word and multi-word ) .
in the cross-validation process , multinomial naive bayes ( mnb ) has shown better results than support vector machines ( svm ) .
in the cross-validation process , multinomial naive bayes ( mnb ) has shown better results than support vector machines ( svm ) as a component for adaboost .
the feature weights are tuned to optimize bleu using the minimum error rate training algorithm .
the feature weights for each system were tuned on development sets using the moses implementation of minimum error rate training .
the expectation maximization algorithm has been widely applied for solving the decipherment problem .
the expectation-maximization algorithm was applied to a wide range of problems .
we start with a bidirectional long short-term memory model that employs pretrained word embeddings .
with reference to this system , we implement a data-driven parser with a neural classifier based on long short-term memory .
corpus normalization and smoothing methods were as described in roark et al .
text normalization and smoothing parameterizations were as presented in roark et al .
we use the maximum entropy model for our classification task .
we use a standard maximum entropy classifier implemented as part of mallet .
in this work , we will be using word graphs .
in this paper , we have considered new input sources for imt .
we used a phrase-based smt model as implemented in the moses toolkit .
our smt system is a phrase-based system based on the moses smt toolkit .
in this paper , we propose a set of additional features , some of which are designed to better capture structural information .
to improve chinese srl , we propose a set of additional features , some of which are designed to better capture structural information .
this matrix is produced from a word representation method such as word2vec .
we use word2vec as the vector representation of the words in tweets .
collobert et al , kalchbrenner et al , and kim use convolutional networks to deal with varying length sequences .
collobert et al use a convolutional neural network over the sequence of word embeddings .
moses is used as a baseline phrase-based smt system .
it is a standard phrasebased smt system built using the moses toolkit .
the weka toolkit was used for all experiments .
for these experiments we use the weka toolkit .
rockt盲schel et al recently proposed a joint model which injects first-order logic into embeddings .
rockt盲schel et al proposed a joint model that injects first-order logic into embeddings .
in this work , we present a method to identify the attitude of participants in an online discussion .
our proposed work is identifying attitudes in sentences that appear in online discussions .
in particular , we use a feature-based lexicalized tree-adjoining grammar , that is derived from an hpsg grammar .
to generate from the kbgen data , we induce a feature-based lexicalised tree adjoining grammar , augmented with a unification-based semantics from the training data .
lerner and petrov present a simple classifier-based preordering approach using the source-side dependency tree .
lerner and petrov train classifiers to predict the permutations of up to 6 tree nodes in the source dependency tree .
coreference resolution is the problem of identifying which mentions ( i.e. , noun phrases ) refer to which real-world entities .
coreference resolution is a task aimed at identifying phrases ( mentions ) referring to the same entity .
we use the stanford pos-tagger and name entity recognizer .
we use the stanford ner system with a standard set of language-independent features .
summarization data sets demonstrate this proposed method outperforms the previous ilp system .
our proposed method yielded better results than the previous state-of-the-art ilp system on different tac data sets .
the translation quality is evaluated by case-insensitive bleu and ter metrics using multeval .
the translation performance was measured using the bleu and the nist mt-eval metrics , and word error rate .
lambert et al did tune on queen , a simplified version of ulc that discards the semantic features and is based on pure lexical features .
lambert et al did tune on queen , a simplified version of ulc that discards the semantic features of ulc and is based on pure lexical similarity .
the ptb parser we use for comparison is the publicly available berkeley parser .
for our parsing experiments , we use the berkeley parser .
this paper considers the problem of modelling temporal frequency profiles of rumours .
this paper introduced the problem of modelling frequency profiles of rumours in social media .
and indeed , the results show the ability of lexicalized surprisal to explain a significant amount of variance in rt data .
the results show that lexicalized surprisal according to both models is a significant predictor of rt , outperforming its unlexicalized counterparts .
projected from the parsed english translation , experiments show that the bilingually-guided method achieves a significant improvement of 28 . 5 % over the unsupervised baseline .
experiments on 5 languages show that the novel strategy significantly outperforms previous unsupervised or bilingually-projected models .
hu and liu used similar lexical network , but they considered not only synonyms but antonyms .
hu and liu use the synonym and antonym relations within linguistic resources .
phrasebased smt models are tuned using minimum error rate training .
each translation model is tuned using mert to maximize bleu .
we trained a 5-grams language model by the srilm toolkit .
for learning language models , we used srilm toolkit .
using recurrent neural networks has become a very common technique for various nlp based tasks like language modeling .
currently , recurrent neural network based models are widely used on natural language processing tasks for excellent performance .
the pun is defined as “ a joke exploiting the different possible meanings of a word or the fact that there are words which sound alike but have different meanings ” ( cite-p-7-1-6 ) .
a pun is a means of expression , the essence of which is in the given context the word or phrase can be understood in two meanings simultaneously ( cite-p-22-3-7 ) .
as reported in , a simple averaging scheme was found to be very competitive to more complex models for representing a sentence vector .
interestingly , as reported in , a simple averaging scheme was found to be very competitive to more complex models for high level semantic tasks despite its simplicity .
kambhatla employs maximum entropy models to combine diverse lexical , syntactic and semantic features derived from the text for relation extraction .
kambhatla leverages lexical , syntactic and semantic features , and feeds them to a maximum entropy model .
for example , or indicate the limited coverage of framenet as one of the main problems of this resource .
for example , indicate the limited coverage of framenet as one of the main problems of this resource .
we used the implementation of random forest in scikitlearn as the classifier .
we used the logistic regression implemented in the scikit-learn library with the default settings .
in this paper , we present an experimental study on solving the answer selection problem .
in this paper , we focus on one of the key subtasks ¨c answer sentence selection .
relation extraction is the task of predicting attributes and relations for entities in a sentence ( zelenko et al. , 2003 ; bunescu and mooney , 2005 ; guodong et al. , 2005 ) .
relation extraction is the task of finding semantic relations between two entities from text .