Datasets:

deepachalapathi

/

parasci_data

like 0

we ran mt experiments using the moses phrase-based translation system---we used moses , a state-of-the-art phrase-based smt model , in decoding

we use the selectfrommodel 4 feature selection method as implemented in scikit-learn---we implemented the algorithms in python using the stochastic gradient descent method for nmf from the scikit-learn package

for each one of the 6 languages which our approach covers , we built a phrase-based machine translation model using the moses toolkit---pitler et al used the data from vadas and curran for a parser applicable on base noun phrases of any length including coordinations

the next section gives an overview of related work---in this section , we summarize related work

accordingly , we have trained 3- and 5-dimensional models for english and german syllable structure---we initialize the embedding layer using embeddings from dedicated word embedding techniques word2vec and glove

for run 2 , we use wapiti , an efficient off-the-shelf linear-chain crf sequence classifier---to train a crf model , we use the wapiti sequence labelling toolkit

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---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

however , we use a large 4-gram lm with modified kneser-ney smoothing , trained with the srilm toolkit , stolcke , 2002 and ldc english gigaword corpora---mikolov et al proposed a faster skip-gram model word2vec 5 which tries to maximize classification of a word based on another word in the same sentence

this paper describes our system participation in the semeval-2017 task 8 ¡®rumoureval : determining rumour veracity and support for rumours¡¯---to calculate language model features , we train traditional n-gram language models with ngram lengths of four and five using the srilm toolkit

later , it has been applied in natural language processing tasks and outperformed traditional models such as bag of words , n-grams and their tfidf variants---performance is measured in terms of perplexity

we used a freely-available pretrained model of 300 dimensions trained on approximately 100 billion words from news articles---we used all post bodies in the unlabeled dataset to train a skip-gram model of 50 dimensions

we use a standard maximum entropy classifier implemented as part of mallet---we utilize a maximum entropy model to design the basic classifier used in active learning for wsd

recently , a series of methods have been developed , which train a classifier for each label , organize the classifiers in a partially ordered structure and take predictions produced by the former classifiers as the latter classifiers ’ features---style methods have been developed , which train a classifier for each label , organize the classifiers in a partially ordered structure and take predictions produced by the former classifiers

for all models , we use l 2 regularization and run 100 epochs of adagrad with early stopping---for the optimization process , we apply the diagonal variant of adagrad with mini-batches

some opinion mining methods in english rely on the english lexicon sentiwordnet for extracting word-level sentiment polarity---in this paper , we propose a novel family of recurrent neural network unit : the context-dependent additive recurrent neural network ( carnn ) that is designed specifically to leverage

most previous studies on meeting summarization have focused on extractive summarization---coreference resolution is a key problem in natural language understanding that still escapes reliable solutions

word sense disambiguation ( wsd ) is the nlp task that consists in selecting the correct sense of a polysemous word in a given context---word sense disambiguation ( wsd ) is the task of identifying the correct sense of an ambiguous word in a given context

the language models are trained on the corresponding target parts of this corpus using the sri language model tool---the target language model is trained by the sri language modeling toolkit on the news monolingual corpus

we implemented a greedy transition-based parser , and used rich contextual features following zhang and nivre---the parsing model is a shift-reduce dependency parser , using the higherorder features from zhang and nivre

in their proposed model , yang et al . ( 2016 ) use bidirectional gru modules to represent segments as well as documents , whereas we use a more efficient cnn encoder to compose words into segment vectors 2 ( i.e. , math-w-3-4-5-220 )---we measure machine translation performance using the bleu metric

in a headed tree , each terminal word can be uniquely labeled with a governing word and grammatical relation---in the introduction , nonce2vec is designed with a view to be an essential component of an incremental concept

we use the 300-dimensional pre-trained word2vec 3 word embeddings and compare the performance with that of glove 4 embeddings---in addition , we utilize the pre-trained word embeddings with 300 dimensions from for initialization

word alignment is a natural language processing task that aims to specify the correspondence between words in two languages ( cite-p-19-1-0 )---recently , distributional features have also been used directly to train classifiers that classify pairs of words as being synonymous or not

under this setting , we compare our method to the spin model described in---our model by construction is similar to approach based on the ising spin model described in

word sense disambiguation ( wsd ) is the task of determining the correct meaning for an ambiguous word from its context---evaluation measures are sufficient to discern simulated from real dialogs

high quality word embeddings have been proven helpful in many nlp tasks---here , we use negative sampling as a speed-up technique

we first sentencesegment the gigaword corpus using the nltk sentence segmenter---more concretely , faruqui and dyer use canonical correlation analysis to project the word embeddings in both languages to a shared vector space

a multiword expression can be defined as a combination of words for which syntactic or semantic properties of the whole expression can not be obtained from its parts---a multiword expression can be defined as any word combination for which the syntactic or semantic properties of the whole expression can not be obtained from its parts

topic models are commonly inferred using either collapsed gibbs sampling rosen-zvi et al 2004 ) or methods based on variational inference---conventional topic models learning approaches are based on gibbs sampling or variational expectation maximization algorithm

marcu and wong propose a model to learn lexical correspondences at the phrase level---marcu and wong proposed a phrase-based context-free joint probability model for lexical mapping

carreras et al and koo et al introduced frameworks for joint learning of phrase-structure and dependency parsers and showed improvements on both tasks for english---carreras et al and rush et al introduced frameworks for joint learning of phrase and dependency structures , and showed improvements on both tasks for english

stolcke et al point out that the use of dialogue acts is a useful first level of analysis for describing discourse structure---stolcke et al used hmms as a general model of discourse with an application to speech acts in conversations

kambhatla leverages lexical , syntactic and semantic features , and feeds them to a maximum entropy model---kambhatla employs maximum entropy models to combine diverse lexical , syntactic and semantic features derived from the text for relation extraction

corry relies on a rich linguistically motivated feature set , which has , however , been manually reduced to 64 features for efficiency reasons---in a language generation system , a content planner typically uses one or more “ plans ” to represent the content to be included in the output

we evaluate our semantic parser on the webques-tions dataset , which contains 5,810 question-answer pairs---for evaluation , we use the webques-tions , a benchmark dataset for qa on freebase

the language model was constructed using the srilm toolkit with interpolated kneser-ney discounting---we estimated 5-gram language models using the sri toolkit with modified kneser-ney smoothing

we also use a 4-gram language model trained using srilm with kneser-ney smoothing---for language model , we used sri language modeling toolkit to train a 4-gram model with modified kneser-ney smoothing

combined with other features , they obtained a test accuracy of 75.29 % on the toefl11 dataset---and combined with customarily used word n-grams , they have high performance in terms of accuracy , when tested on the toefl11 corpus

our model is a structured conditional random field---we primarily compared our model with conditional random fields

luong et al , 2013 ) utilized recursive neural networks in which inputs are morphemes of words---for which the automatic evaluation metrics proposed to date for machine translation and automatic summarization can be seen as particular instances

given context vectors , lin and pantel used a symmetric similarity metric to find candidate paraphrases---lin and pantel use a standard monolingual corpus to generate paraphrases , based on dependancy graphs and distributional similarity

klein and manning , for example , show that the performance of an unlexicalised model can be substantially improved by splitting the existing symbols down into finer categories---klein and manning show that much of the gain in statistical parsing using lexicalized models comes from the use of a small set of function words

we estimate a 5-gram language model using interpolated kneser-ney discounting with srilm---we use the sri language model toolkit to train a 5-gram model with modified kneser-ney smoothing on the target-side training corpus

in a second baseline model , we also incorporate 300-dimensional glove word embeddings trained on wikipedia and the gigaword corpus---we use 300-dimensional glove vectors trained on 6b common crawl corpus as word embeddings , setting the embeddings of outof-vocabulary words to zero

grammar induction is the task of inducing high-level rules for application of grammars in spoken dialogue systems---the language model is a trigram-based backoff language model with kneser-ney smoothing , computed using srilm and trained on the same training data as the translation model

a 4-gram language model is trained on the xinhua portion of the gigaword corpus with the srilm toolkit---a trigram language model with modified kneser-ney discounting and interpolation was used as produced by the srilm toolkit

it is worth noting that the morpheme feature is employed to better represent the compositional semantics inside chinese words---we employ a new feature ( morpheme feature ) which is particularly appropriate for chinese

to solve this task we use a multi-class support vector machine as implemented in the liblinear library---we train a linear support vector machine classifier using the efficient liblinear package

twitter is a popular microblogging service , which , among other things , is used for knowledge sharing among friends and peers---twitter is a well-known social network service that allows users to post short 140 character status update which is called “ tweet ”

we also use a 4-gram language model trained using srilm with kneser-ney smoothing---we use srilm toolkit to build a 5-gram language model with modified kneser-ney smoothing

to integrate their strengths , in this paper , we propose a forest-based tree sequence to string translation model---in this paper , we propose a forest-based tree sequence to string model , which is designed to integrate the strengths of the forest-based and the tree

sarcasm is defined as ‘ the use of irony to mock or convey contempt ’ 1---sarcasm is defined as ‘ a cutting , often ironic remark intended to express contempt or ridicule ’ 1

figure 6 shows that our approaches consistently outperform the baseline and the state-of-the-art methods with diverse feature sparsity degrees---features demonstrate that our low-rank matrix completion approach significantly outperforms the baseline and the state-of-the-art methods

both corpora were extracted from the open parallel corpus opus---the central component of our non-parametric bayesian model are pitman-yor processes , which are a generalization of the dirichlet processes

different dialogue act labeling standards and datasets have been provided , including switchboard-damsl , icsi-mrda and ami---hermjakob implemented a shift-reduce parser for korean trained on very limited data , and sarkar and han used an earlier version of the treebank to train a lexicalized tree adjoining grammar

we apply our model to the english portion of the conll 2012 shared task data , which is derived from the ontonotes corpus---we train and evaluate our model on the english corpus of the conll-2012 shared task

we used the target side of the parallel corpus and the srilm toolkit to train a 5-gram language model---for the tree-based system , we applied a 4-gram language model with kneserney smoothing using srilm toolkit trained on the whole monolingual corpus

lexical simplification is a subtask of text simplification ( cite-p-16-3-3 ) concerned with replacing words or short phrases by simpler variants in a context aware fashion ( generally synonyms ) , which can be understood by a wider range of readers---lexical simplification is a specific case of lexical substitution where the complex words in a sentence are replaced with simpler words

we use srilm to build 5-gram language models with modified kneser-ney smoothing---this type of features are based on a trigram model with kneser-ney smoothing

the system is an almost delexicalized parser which does not need training data to analyze romance languages---and exhibits stable performance across languages

sun and xu uses punctuation information as discrete feature in a sequence labeling framework , which shows improvement compared to the pure sequence labeling approach---chang and han , sun and xu used rich statistical information as discrete features in a sequence labeling framework

latent dirichlet allocation is a generative model that overcomes some of the limitations of plsi by using a dirichlet prior on the topic distribution---latent dirichlet allocation is a generative probabilistic topic model where documents are represented as random mixtures over latent topics , characterized by a distribution over words

kalchbrenner et al proposed a dynamic convolution neural network with multiple layers of convolution and k-max pooling to model a sentence---kalchbrenner et al developed a cnnbased model that can be used for sentence modelling problems

sentiment analysis is a fundamental problem aiming to give a machine the ability to understand the emotions and opinions expressed in a written text---we use pre-trained 50-dimensional word embeddings vector from glove

1a bunsetsu is a common unit when syntactic structures in japanese are discussed---a bunsetsu consists of one independent word and zero or more ancillary words

the berkeley parser is an efficient and effective parser that introduces latent annotations to refine syntactic categories to learn better pcfg grammars---the berkeley parser is an efficient and effective parser that introduces latent annotations to learn high accurate context-free grammars directly from a treebank

in this work , we propose a multi-space variational encoder-decoder framework for labeled sequence transduction problem---in this work , we propose a new framework for labeled sequence transduction problems : multi-space variational encoder-decoders

finally , we explain how to let our model additionally learn the language ’ s canonical word order---additionally letting our model learn the language ’ s canonical word order improves its performance and leads to the highest semantic parsing

in this paper , we have discussed possibilities to translate via pivot languages on the character level---in this paper we investigate the use of character-level translation models to support the translation from and to under-resourced languages

this capability is very desirable as shown by the success of the rule-based deterministic approach of raghunathan et al in the conll shared task 2011---in fact , the rule-based system of raghunathan et al exhibited the top score in the recent conll evaluation

we have shown co-training to be a promising approach for predicting emotions with spoken dialogue data---we investigate the applicability of co-training to train classifiers that predict emotions in spoken dialogues

unfortunately , this approach is difficult to utilize because it requires multiple segmenters that behave differently on the same input---however , much of this work has relied on multiple segmenters that perform differently on the same input

we have not yet succeeded , however , in combining the benefits of both prosody and the hbm---and while we have not been able to usefully employ both prosody and the hbm technique together , our hbm is competitive

semantic parsing is the task of transducing natural language ( nl ) utterances into formal meaning representations ( mrs ) , commonly represented as tree structures---semantic parsing is the task of mapping natural language sentences to a formal representation of meaning

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

we present a graph-based semi-supervised learning for the question-answering ( qa ) task for ranking candidate sentences---salehi et al show that word embeddings are more accurate in predicting compositionality than a simplistic count-based dsm

different types of architectures such as feedforward neural networks and recurrent neural networks have since been used for language modeling---recurrent neural network architectures have proven to be well suited for many natural language generation tasks

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

stance detection is the task of automatically determining from the text whether the author of the text is in favor of , against , or neutral towards a proposition or target---we used moses as the implementation of the baseline smt systems

based on uima , it allows for efficient parallel processing of large volumes of text---distribution allows for straightforward high-performance nlp processing

lexical simplification is a popular task in natural language processing and it was the topic of a successful semeval task in 2012 ( cite-p-14-1-9 )---we used the pre-trained word embeddings that were learned using the word2vec toolkit on google news dataset

gildea and jurafsky are the only ones applying selectional preferences in a real srl task---gildea and jurafsky is the only one applying selectional preferences in a real srl task

in this paper , we propose a novel approach for disfluency detection---recently , inversion transduction grammars , namely itg , have been used to constrain the search space for word alignment

to train the link embeddings , we use the speedy , skip-gram neural language model of mikolov et al via their toolkit word2vec---we use a popular word2vec neural language model to learn the word embeddings on an unsupervised tweet corpus

dependency parsing is the task of building dependency links between words in a sentence , which has recently gained a wide interest in the natural language processing community---therefore , dependency parsing is a potential “ sweet spot ” that deserves investigation

this single endto-end nmt model outperforms the best conventional smt system ( cite-p-20-1-5 ) and achieves a state-of-the-art performance---single nmt model achieves state-of-the-art performance and outperforms the best conventional model

moore and lewis calculated the difference of the cross entropy values for a given sentence , based on language models from the source domain and the target domain---another approach is taken by , where , based on source and target language models , the authors calculated the difference of the cross-entropy values for a given sentence

coreference resolution is a field in which major progress has been made in the last decade---coreference resolution is the task of determining which mentions in a text refer to the same entity

mikolov et al observed a strong similarity of the geometric arrangements of corresponding concepts between the vector spaces of different languages , and suggested that a crosslingual mapping between the two vector spaces is technically plausible---and more importantly , if vectors learned for languages are manually rotated , mikolov et al observed that languages share similar geometric arrangements in vector spaces

in this work , we tackle the challenge of extracting bursty phrases without any restriction of forms---in this work , we aim to accurately and exhaustively extract bursty phrases of arbitrary forms

in this paper , we propose a constrained word lattice to combine smt and tm at phrase-level---in this paper , we propose using a constrained word lattice , which encodes input phrases and tm constraints

the availability of large document-summary corpora have opened up new possibilities for using statistical text generation techniques for abstractive summarization---availability of large document-summary corpora , as we discuss in section 3 , has opened up new possibilities for applying statistical text generation approaches to summarization

the skip-gram and continuous bag-of-words models of mikolov et al propose a simple single-layer architecture based on the inner product between two word vectors---mikolov et al further proposed continuous bagof-words and skip-gram models , which use a simple single-layer architecture based on inner product between two word vectors

for this , we utilize the publicly available glove 1 word embeddings , specifically ones trained on the common crawl dataset---we use 300 dimensional glove embeddings trained on the common crawl 840b tokens dataset , which remain fixed during training

in this paper we addressed the problem of recommending questions from large archives of community question answering data based on users¡¯ information needs---to address the aforementioned problems of the vsm model , the sentiment vector space model ( s-vsm ) is proposed

user simulation is frequently used to train statistical dialog managers for task-oriented domains---dialogs has become vital for training statistical dialog managers in task-oriented domains

for minimum error rate tuning , we use nist mt-02 as the development set for the translation task---we perform minimum-error-rate training to tune the feature weights of the translation model to maximize the bleu score on development set

with the svm reranker , we obtain a significant improvement in bleu scores over white & rajkumar¡¯s averaged perceptron model on both development and test data---with the svm reranker , we obtain a significant improvement in bleu scores over white & rajkumar ¡¯ s averaged perceptron model

the word embeddings are initialized using the pre-trained glove , and the embedding size is 300---and we used a graph kernel instead of a sequence kernel to measure the similarity between pairs of documents

tuning is performed to maximize bleu score using minimum error rate training---the weights used during the reranking are tuned using the minimum error rate training algorithm