tasksource/scinli · Datasets at Hugging Face

id stringlengths 7 12	sentence1 stringlengths 6 1.27k	sentence2 stringlengths 6 926	label stringclasses 4 values
train_3700	(2003), word embeddings aim at representing, i.e., embedding, the ideal semantic space of words in a real-valued continuous vector space.	to traditional distributional techniques, such as Latent Semantic Analysis (Landauer and Dutnais, 1997, LSA) and Latent Dirichlet Allocation (Blei et al., 2003, LDA), Bengio et al.	contrasting
train_3701	(2014), which leverage embeddings for supervised (the former three) and knowledge-based (the latter) WSD.	to our knowledge, no previous work has investigated methods for integrating word embeddings in WSD and the role that different training parameters can play.	contrasting
train_3702	Among combination strategies, concatenation and average produced the smallest gain and did not benefit from embeddings of higher dimensionality.	the other two strategies, i.e., fractional and exponential decay, showed improved performance with the increase in the size of the employed embeddings, irrespective of the WSD features.	contrasting
train_3703	The more complex Impatient Reader computes attention over the document after reading every word of the query.	empirical evaluation has shown that both models perform almost identically on the CNN and Daily Mail datasets.	contrasting
train_3704	Figure 7a shows that the accuracy significantly drops as the correct answer gets less and less frequent in the document compared to other candidate answers.	the correct answer is likely to occur frequently (Fig.	contrasting
train_3705	More recently, methods that jointly infer the opinion entity and relation extraction tasks (e.g., using Integer Linear Programming (ILP)) have been introduced (Choi et al., 2006;Yang and Cardie, 2013) and show that the existence of opinion relations provides clues for the identification of opinion entities and vice-versa, and thus results in better performance than a pipelined approach.	the success of these methods depends critically on the availability of opinion lexicons, dependency parsers, named-entity taggers, etc.	contrasting
train_3706	They are capable of modeling sequences of arbitrary length by repetitive application of a recurrent unit along the tokens in the sequence.	recurrent neural networks are known to have several disadvantages like the problem of vanishing and exploding gradients.	contrasting
train_3707	For the first hidden layer, the computation proceeds similar to that described in Section 3.1.	for higher hidden layers i the input to the memory block is the hidden state and memory cell from the previous layer i − 1 instead of the input vector representation.	contrasting
train_3708	With respect to holders and targets, we find that our SLL model performs similar to the CRF baseline.	the SLL+RLL model outperforms CRF baseline.	contrasting
train_3709	Our sequential LSTM model is able to learn these relations thus validating that LSTMs can model long-term dependencies.	for IS-FROM relations, we find that our recall is lower than the ILP-based joint model.	contrasting
train_3710	Averaging the embeddings of words in a sentence has proven to be a surprisingly successful and efficient way of obtaining sentence embeddings.	word embeddings trained with the methods currently available are not optimized for the task of sentence representation, and, thus, likely to be suboptimal.	contrasting
train_3711	Speedups might of course be gained for all methods by presenting the sentences in batches to the models, by computing sentence representations in parallel and by running code on a GPU.	as we are interested in the differences between the systems, we run the most simple and straightforward scenario.	contrasting
train_3712	It is beyond the scope of this paper to provide a comprehensive analysis of all supervised methods using word or sentence embeddings and the effect Siamese CBOW would have on them.	it would be interesting to see how Siamese CBOW embeddings would affect results in supervised tasks.	contrasting
train_3713	With negative mappings, null maps to all:e, while each proper noun maps to its proper predicate minus all:e. There is a lot of work in semantic parsing that tackles the GeoQuery dataset (Zelle and Mooney, 1996;Zettlemoyer and Collins, 2005;Wong and Mooney, 2007;Kwiatkowski et al., 2010;Liang et al., 2011), and the state-of-the-art is 91.1% precision and recall (Liang et al., 2011).	none of these methods can guarantee 100% precision, and they perform more feature engineering, so these numbers are not quite comparable.	contrasting
train_3714	Another avenue for providing user confidence is probabilistic calibration (Platt, 1999), which has been explored more recently for structured prediction (Kuleshov and Liang, 2015).	these methods do not guarantee precision for any training set and test input.	contrasting
train_3715	Although some RNN models showed little performance gains over the SVM baselines only with doc2vec features, they were even worse than the CRF model with the same features.	the RCNN models connecting the results of CNNs to the RNNs contributed to performance improvements not only from the baselines, but also from the CNN models.	contrasting
train_3716	While the CNN model tended to have a higher coverage in segmentation than the others, the LRCN model produced more precise decisions to recognize the boundaries on the strength of the consideration of conversational coherences in dialogue history sequences.	the segmentation performances even with the best models were still very limited especially for inter-categorical transitions.	contrasting
train_3717	Secondly, the use of external knowledge could be a key to success in dialogue topic tracking, as proved in the previous studies.	this work only takes internal dialogue information into account for making decisions.	contrasting
train_3718	Some of these studies also demonstrated that using L1specific properties, such as the error patterns of speakers of a given L1 or L1-induced paraphrases, improves the performance of automatic error correction in non-native writing.	neither of the approaches has constructed a semantic model from L1 data and systematically studied the effects of its transfer onto L2.	contrasting
train_3719	The results of these studies suggest that L1 is a valuable source of information in EDC.	all these works use isolated translational equivalents and focus on error correction only.	contrasting
train_3720	Adding the L1 lexicosemantic features has only a minor effect on accuracy and precision, and a more pronounced effect on recall.	when we test the system on one particular related L1 (Table 8) we observe the opposite effect: with the exception of ES subj data, precision and accuracy improve, suggesting that the error detection system using L1-induced information identifies errors more precisely.	contrasting
train_3721	For distractors considered "Plausible" by both experts, 63.6% were selected by the learners.	for those considered "Obviously wrong" by both experts, only 11.8% attracted any learner.	contrasting
train_3722	Limsopatham and Collier (2015a) leveraged translations between the informal language used in social media and the formal language used in the description of medical concepts in an ontology.	we argue that effective concept normalisation requires a system to take into account the semantics of social media messages and medical concepts.	contrasting
train_3723	We observe that for TwADR-S and TwADR-L datasets, which are smaller datasets (dataset size of 201 and 1,436, respectively), a better performance can be achieved if the model is not allowed to update the embeddings of the input phrases.	for the AskAPatient dataset (dataset size of 8,662), allowing the model to update the embeddings results in a significantly (paired t-test, p < 0.05) better performance.	contrasting
train_3724	We neglected such time-related phrases when defining C 0 .	the frequencies of the listed n-grams in the business group are much higher than those in the individual group.	contrasting
train_3725	For example, English POS taggers benefit from carefully designed word spelling features; orthographic features and external resources such as gazetteers are widely used in NER.	such task-specific knowledge is costly to develop (Ma and Xia, 2014), making sequence labeling models difficult to adapt to new tasks or new domains.	contrasting
train_3726	Several RNN-based neural network models have been proposed to solve sequence labeling tasks like speech recognition (Graves et al., 2013), POS tagging and NER (Chiu and Nichols, 2015;Hu et al., 2016), achieving competitive performance against traditional models.	even systems that have utilized distributed representations as inputs have used these to augment, rather than replace, hand-crafted features (e.g.	contrasting
train_3727	For many sequence labeling tasks it is beneficial to have access to both past (left) and future (right) contexts.	the LSTM's hidden state h t takes information only from past, knowing nothing about the future.	contrasting
train_3728	The RNNLM could be used in a KNN-style approach, where it associates each example response with its individual topic vector, using L(q) as a distance metric.	this is computationally infeasible since computing L(q) is significantly more expensive than cosine distance and the previously mentioned scalability would be lost.	contrasting
train_3729	SC-EX2, which relates to similar issues, is often confused with it.	sC-EX3 is rarely confused with sC-EX1 as it is about non-personal events on a larger scale.	contrasting
train_3730	A limitation of both the standard and proposed approach is that if a new question is created by the test-makers, then it will be necessary to collect example responses before it can be widely deployed.	since the system can be trained on ASR transcriptions, the example responses do not need to be hand-transcribed.	contrasting
train_3731	This makes automatic detection of sarcasm an important problem.	it has been quite difficult to solve such a problem with traditional NLP tools and techniques.	contrasting
train_3732	(2016) for modeling sarcasm understandability of readers.	as far as we know, these features are being introduced in NLP tasks like textual sarcasm detection for the first time.	contrasting
train_3733	(2013) conducted a corpus analysis and found certain similarity between Penn Discourse TreeBank relations (Prasad et al., 2008) and argumentation schemes (Walton et al., 2008).	they did not discuss how such similarity could be applied to argument mining.	contrasting
train_3734	Argument words (e.g., 'believe', 'opinion') signal the argumentative content and are commonly used across different topics.	domain words are specific terminologies commonly used within the topic (e.g., 'art', 'education').	contrasting
train_3735	If you know what labels you want, active learning can reduce the number of labeled documents needed.	establishing the label set remains difficult.	contrasting
train_3736	In LA and LR, the preference function directly chooses a document and directs the user to it.	u TA d and u TR d are topic dependent.	contrasting
train_3737	It is designed to detect error patterns from a fixed window of 7 words, which is large enough to not require the use of more advanced composition functions.	the performance of the bidirectional recurrent network (Bi-RNN) is somewhat lower, especially on the test set.	contrasting
train_3738	It is worth noting that the full Bi-LSTM has been trained on more data than the other CoNLL contestants.	as the shared task systems were not restricted to the NUCLE training set, all the submissions also used differing amounts of training data from various sources.	contrasting
train_3739	Through the availability of large annotated resources, such as PropBank (Palmer et al., 2005), statistical models based on such features achieve high accuracy.	results often fall short when the input to be labeled involves instances of linguistic phenomena that are relevant for the labeling decision but appear infrequently at training time.	contrasting
train_3740	Analogously to how word embeddings are computed, the simplest way to embed paths would be to represent each sequence as a one-hot vector.	this is suboptimal for two reasons: Firstly, we expect only a subset of dependency paths to be attested frequently in our data and therefore many paths will be too sparse to learn reliable embeddings for them.	contrasting
train_3741	The prior studies attempt to predict how strongly users are currently engaged in dialogues with systems.	the goal of this study is to predict how strongly users will be engaged with intelligent assistants in the future.	contrasting
train_3742	chat Voice Assist can give greetings to, have conversations with, and play games with users (e.g., V2 and V6).	to device operations for accomplishing certain tasks, these functions are offered for fun or for facilitating smooth communication.	contrasting
train_3743	In the trigger classification stage, some previous approaches (Grishman et al., 2005;Ji and Grishman, 2008;Liao and Grishman, 2010;Huang and Riloff, 2012) use patterns to decide the types of event triggers.	pattern-based approaches suffer from low recall since real world events usually have a large variety of representations.	contrasting
train_3744	DMCNN is mainly used to model contextual features.	dMCNN still does not consider argument-argument interactions.	contrasting
train_3745	We can see that RBPB with both plain feature and pattern feature can gain Table 3: The effect (F 1 value) of pattern feature much better performance than with two kinds of features alone.	our approach is just a pipeline approach which suffers from error propagation and the argument performance may not affect the trigger too much.	contrasting
train_3746	We find that JET depends too much on event patterns for event type priori and JET considers each candidate argument separately.	patterns cannot cover all events and the relationship between candidate arguments may help when identifying arguments.	contrasting
train_3747	Therefore, we develop an event type classifier to assign the event type, using both pattern matching information and other features, which gives our system the capability to deal with failed match cases when using patterns alone.	we train a maximum entropy classifier to predict the relationship between candidate arguments.	contrasting
train_3748	In summary, by using the event type classifier and the regularization method, we have achieved a good performance in which the trigger classification is comparable to state-of-theart methods, and the argument identification & classification performance is significantly better than state-of-the-art methods.	we only use sentence-level features and our method is a pipelined approach.	contrasting
train_3749	With self-training, we successfully exploit unlabeled data to improve over ENT by 15% F-score on the newswire domain, and 13% F-score on clinical data.	our active learning experiments demonstrate that we can match (and even beat) ENT using only 6.6% of the training data in the clinical domain, and only 5.8% of the training data in the newswire domain.	contrasting
train_3750	On RTE-2 their method achieves accuracy improvements of upto 10%.	hickl and Bensley (2007) achieves only a 1% accuracy improvement on RTE-3 using the same method, suggesting that it is not always as beneficial.	contrasting
train_3751	A token in one text passage that matches exactly, characterfor-character, with a token in another text passage would be considered a term match by this simple term matcher.	these term matchers could be more sophisticated and match pairs of terms that are synonyms, or paraphrases, or Table 3.	contrasting
train_3752	Our self-training system substantially improved over ENT, achieving an F-score gain of 15% on newswire and 13% on clinical, using only additional unlabeled data.	our active learning experiments demonstrated that we could match (and even beat) the baseline ENT system with only 6.6% of the training data in the clinical domain, and only 5.8% of the training data in the newswire domain.	contrasting
train_3753	Existing multilingual topic models consider only document-level alignments.	most documents are hierarchically structured, i.e., a document comprises segments (e.g., sections and paragraphs) that can be aligned across languages.	contrasting
train_3754	BiSTM leverages existing segments from a given segmentation.	a segmentation is not always given, and a given segmentation might not be optimal for statistical modeling.	contrasting
train_3755	This seems to be reasonable, because some original sections include multiple topics.	tables 2 and 3 show that inferred boundaries do not work better than section boundaries.	contrasting
train_3756	Orthogonal to the formal semantics of DCS, distributional vector representations are useful in capturing lexical semantics of words (Turney and Pantel, 2010;Levy et al., 2015), and progress is made in combining the word vectors to form meanings of phrases/sentences (Mitchell and Lapata, 2010; Baroni and Zamparelli, 2010;Grefenstette and Sadrzadeh, 2011;Socher et al., 2012;Paperno et al., 2014;Hashimoto et al., 2014).	less effort is devoted to finding a link between vector-based compositions and the composition operations in any formal semantics.	contrasting
train_3757	It suggests regularities in the semantic space, especially because orthogonal matrices preserve cosine similarity -if M N is orthogonal, two words x, y and their projections π N (x), π N (y) will have the same similarity measure, which is semantically reasonable.	matrices trained by vecUD are only orthogonal for three UD relations, namely conj, dep and appos.	contrasting
train_3758	(2014) and the entanglement model in Kartsaklis and Sadrzadeh (2014).	these models do not show particular advantages on other datasets.	contrasting
train_3759	Results As shown in Table 7, vecDCS scores better than the N-gram model and demonstrates promising performance.	to our surprise, "no matrix" shows an even better result which is the new state-of-the-art.	contrasting
train_3760	Logic and Distributional Semantics Logic is necessary for implementing the functional aspects of meaning and organizing knowledge in a structured and unambiguous way.	distributional semantics provides an elegant methodology for assessing semantic similarity and is well suited for learning from data.	contrasting
train_3761	(2014), in exploring vector calculations that realize logic operations.	the previous works did not specify how to integrate contextual distributional information, which is necessary for calculating semantic similarity.	contrasting
train_3762	word embedding) in the answer directly, and then we use RNN to model the attentive word sequence.	this model attends a sentence word by word which may ignore the relation between words.	contrasting
train_3763	In the example given above, the verb corresponds quite closely to the desired target relation.	in the wild, we encounter a multitude of different ways of expressing the same kind of relationship.	contrasting
train_3764	(2014) proposed a deep convolutional neural network with softmax classification, extracting lexical and sentence level features.	these approaches still depend on additional features from lexical resources and NLP toolkits.	contrasting
train_3765	In the training phase, we could probably find that every two entities connected by this relation are also connected by the path editor/film −1 , and hence assign an extremely high weight to it.	5 in the testing phase, for any entity pair (x, y) such that (y, editor/film, x) has not been encoded, we might not even find that path and hence could always make a negative prediction.	contrasting
train_3766	Website privacy policies are often ignored by Internet users, because these documents tend to be long and difficult to understand.	the significance of privacy policies greatly exceeds the attention paid to them: these documents are binding legal agreements between website operators and their users, and their opaqueness is a challenge not only to Internet users but also to policy regulators.	contrasting
train_3767	Legal text is a familiar domain for natural language processing, and the legal community has demonstrated some reciprocal interest (Mahler, 2015).	the scale of the problem and its significance-i.e., to virtually any Internet user, as well as to website operators and policy regulators-distinguishes it and provides immense motivation (Sadeh et al., 2013).	contrasting
train_3768	Other approaches leverage topic mod-eling (Chundi and Subramaniam, 2014;Stamey and Rossi, 2009) or sequence alignment techniques to analyze privacy policies or identify similar policy sections and paragraphs.	the complexity and vagueness of privacy policies makes it difficult to automatically extract complex data practices from privacy policies without substantial gold standard data.	contrasting
train_3769	The search is based on a so-called "floating parser" (Pasupat and Liang, 2015), a modification of a standard chart-parser, which is able to guide the search based on the similarity features.	our approach does not search among the derivations for the one that maximizes a match with the NL, but instead directly tries to predict a decision sequence that can be mapped to the LF.	contrasting
train_3770	number of rules) is larger than that of LF or CF.	dS is unique in allowing us to easily validate grammatical constraints.	contrasting
train_3771	gives small probability to ungrammatical choices, a property not shared by DSP-CL.	a more complete understanding of the difference will need more investigation.	contrasting
train_3772	Word embeddings-distributed representations of words-in deep learning are beneficial for many tasks in NLP.	different embedding sets vary greatly in quality and characteristics of the captured information.	contrasting
train_3773	Each embedding set contains information about only a part of the overall vocabulary.	it can predict what the remaining part should look like by comparing words it knows with the information other embedding sets provide about these words.	contrasting
train_3774	(i) Not all individual embedding sets are beneficial in this task; e.g., HLBL embeddings make FLORS perform worse in 11 out of 12 cases.	(ii) in most cases, embeddings improve system performance, which is consistent with prior work on using embeddings for this type of task (Xiao and Guo, 2013;Yang and Eisenstein, 2014;Tsuboi, 2014).	contrasting
train_3775	As a concrete example we collect live text data and corresponding news reports for football (called soccer more often in the United States) games and conduct our study thereby.	our methods and discussions made in this paper can be trivially adapted to other types of sports games as well.	contrasting
train_3776	This is used for traditional summarization problems (Kulesza and Taskar, 2011).	the problem for constructing sports news from live broadcast script is rather different.	contrasting
train_3777	The comparison between RF and RF+DPP shows the effectiveness of our sentence selection strategy.	the increase is still limited 12 .	contrasting
train_3778	All the sentences are describing a key scoring event.	none of them were selected to construct the news because our LTR model assigns low scores for these short sentences.	contrasting
train_3779	In our experiments, methods with higher ROUGE scores can indeed achieve better coverage of important units such as events, as shown in pyramid scores in Table 2.	we can also observe from Table 5 that automatic metrics currently cannot reflect readability factors very well.	contrasting
train_3780	(2010) introduced a joint rule selection model for hierarchical phrase-based translation, which also approximated the rule selection problem by a binary classification problem like our approach.	these two models adopted linear classifiers similar to those used in the MERS model , which suffers more from the data sparsity problem compared to the CSRS model.	contrasting
train_3781	In traditional linear models, combining scores can be implemented by including low-order features.	for neural models, this is not that straightforward because of nonlinearity.	contrasting
train_3782	The neural rerankers are capable of capturing global syntax features across the tree.	the most non-local neural parser with LSTM cannot exploit global features.	contrasting
train_3783	These methods aim to learn common feature representations for different languages.	most of the current researches only focus on bilingual word embedding.	contrasting
train_3784	grSemi-CRFs use much fewer input features and most of them are task-insensitive 13 .	grSemi-CRFs achieve almost the same performance, sometimes even better.	contrasting
train_3785	For text chunking, grSemi-CRF outperforms all reported supervised models, except JESS-CM (Suzuki and Isozaki, 2008), a semi-supervised model using giga-word scale unlabeled data in training 14 .	the performance of our grSemi-CRF (95.01%) is very close to that of JESS-CM (95.15%).	contrasting
train_3786	Summarization of event descriptions can be improved by selecting causally motivated sentences.	causality is frequently expressed implicitly, which requires world knowledge and inference.	contrasting
train_3787	The explosion caused people to evacuate the building.	the words can not be substituted in the following sentence: The baker made a cake.	contrasting
train_3788	Similar to the FrameNet features, we split the example into three sections.	we also consider the dependency parse of the data.	contrasting
train_3789	VerbNet VerbNet is a resource devoted to storing information for verbs (Kipper et al, 2000).	to WordNet, VerbNet provides a more fine-grained description of events while focusing less on polysemy.	contrasting
train_3790	Modeling relation paths has offered significant gains in embedding models for knowledge base (KB) completion.	enumerating paths between two entities is very expensive, and existing approaches typically resort to approximation with a sampled subset.	contrasting
train_3791	Therefore prior work selected a subset of paths through sampling and pruning (Neelakantan et al., 2015;Lin et al., 2015).	the properties of the BILINEAR composition function for path representation enable us to incrementally build the sums of all path representations exactly, using dynamic programming.	contrasting
train_3792	In either case, the accuracy difference between ALL-PATHS and PRUNED-PATHS is small, and ALL-PATHS mainly gains in efficiency.	when nodes are modeled, the compositional learning approach gains in accuracy as well, especially when text is jointly embedded.	contrasting
train_3793	For the DNN models, we add the penalty term λ θ 2 to the losses, where λ is the regularization coefficient and θ contains all other parameters.	for the bilinear models we regularize the relation matrices M R toward the identity matrix instead of all zeroes, adding the following to the loss: where I r is the r × r identity matrix, the summation is performed over all relations R, and θ represents all other parameters.	contrasting
train_3794	Figure 6 shows how the models perform as the number of utterances per example varies.	when the search space is small (fewer number of utterances), Model B outperforms or is competitive with Model C. as the search space increases (tighter computational constraints), Model C does increasingly better.	contrasting
train_3795	Even though Model A is unable to learn anything with beam size < 240.	for beam sizes larger than 240, Model A attains 100% accuracy.	contrasting
train_3796	Recall that projecting from Model A to C creates a more computationally tractable model at the cost of expressivity.	this is because Model C used a linear model.	contrasting
train_3797	Tree-structured neural networks exploit valuable syntactic parse information as they interpret the meanings of sentences.	they suffer from two key technical problems that make them slow and unwieldy for large-scale NLP tasks: they usually operate on parsed sentences and they do not directly support batched computation.	contrasting
train_3798	Our results show that a model that uses tree-structured composition fully (SPINN) outperforms one which uses it only partially (tree-based CNN), which in turn outperforms one which does not use it at all (RNN).	the full SPINN performed moderately well at reproducing the Stanford Parser's parses of the SNLI data at a transition-by-transition level, with 92.4% accuracy at test time.3 its transi-3Note that this is scoring the model against automatic tion prediction errors are fairly evenly distributed across sentences, and most sentences were assigned partially invalid transition sequences that either left a few words out of the final representation or incorporated a few padding tokens into the final representation.	contrasting
train_3799	For SG model, the max selectivity of the model neurons is only just above 0.7.	the context-based distributed models showed strong selective activation towards country names in Indonesian.	contrasting

train_3700

(2003), word embeddings aim at representing, i.e., embedding, the ideal semantic space of words in a real-valued continuous vector space.

to traditional distributional techniques, such as Latent Semantic Analysis (Landauer and Dutnais, 1997, LSA) and Latent Dirichlet Allocation (Blei et al., 2003, LDA), Bengio et al.

contrasting

train_3701

(2014), which leverage embeddings for supervised (the former three) and knowledge-based (the latter) WSD.

to our knowledge, no previous work has investigated methods for integrating word embeddings in WSD and the role that different training parameters can play.

contrasting

train_3702

Among combination strategies, concatenation and average produced the smallest gain and did not benefit from embeddings of higher dimensionality.

the other two strategies, i.e., fractional and exponential decay, showed improved performance with the increase in the size of the employed embeddings, irrespective of the WSD features.

contrasting

train_3703

The more complex Impatient Reader computes attention over the document after reading every word of the query.

empirical evaluation has shown that both models perform almost identically on the CNN and Daily Mail datasets.

contrasting

train_3704

Figure 7a shows that the accuracy significantly drops as the correct answer gets less and less frequent in the document compared to other candidate answers.

the correct answer is likely to occur frequently (Fig.

contrasting

train_3705

More recently, methods that jointly infer the opinion entity and relation extraction tasks (e.g., using Integer Linear Programming (ILP)) have been introduced (Choi et al., 2006;Yang and Cardie, 2013) and show that the existence of opinion relations provides clues for the identification of opinion entities and vice-versa, and thus results in better performance than a pipelined approach.

the success of these methods depends critically on the availability of opinion lexicons, dependency parsers, named-entity taggers, etc.

contrasting

train_3706

They are capable of modeling sequences of arbitrary length by repetitive application of a recurrent unit along the tokens in the sequence.

recurrent neural networks are known to have several disadvantages like the problem of vanishing and exploding gradients.

contrasting

train_3707

For the first hidden layer, the computation proceeds similar to that described in Section 3.1.

for higher hidden layers i the input to the memory block is the hidden state and memory cell from the previous layer i − 1 instead of the input vector representation.

contrasting

train_3708

With respect to holders and targets, we find that our SLL model performs similar to the CRF baseline.

the SLL+RLL model outperforms CRF baseline.

contrasting

train_3709

Our sequential LSTM model is able to learn these relations thus validating that LSTMs can model long-term dependencies.

for IS-FROM relations, we find that our recall is lower than the ILP-based joint model.

contrasting

train_3710

Averaging the embeddings of words in a sentence has proven to be a surprisingly successful and efficient way of obtaining sentence embeddings.

word embeddings trained with the methods currently available are not optimized for the task of sentence representation, and, thus, likely to be suboptimal.

contrasting

train_3711

Speedups might of course be gained for all methods by presenting the sentences in batches to the models, by computing sentence representations in parallel and by running code on a GPU.

as we are interested in the differences between the systems, we run the most simple and straightforward scenario.

contrasting

train_3712

It is beyond the scope of this paper to provide a comprehensive analysis of all supervised methods using word or sentence embeddings and the effect Siamese CBOW would have on them.

it would be interesting to see how Siamese CBOW embeddings would affect results in supervised tasks.

contrasting

train_3713

With negative mappings, null maps to all:e, while each proper noun maps to its proper predicate minus all:e. There is a lot of work in semantic parsing that tackles the GeoQuery dataset (Zelle and Mooney, 1996;Zettlemoyer and Collins, 2005;Wong and Mooney, 2007;Kwiatkowski et al., 2010;Liang et al., 2011), and the state-of-the-art is 91.1% precision and recall (Liang et al., 2011).

none of these methods can guarantee 100% precision, and they perform more feature engineering, so these numbers are not quite comparable.

contrasting

train_3714

Another avenue for providing user confidence is probabilistic calibration (Platt, 1999), which has been explored more recently for structured prediction (Kuleshov and Liang, 2015).

these methods do not guarantee precision for any training set and test input.

contrasting

train_3715

Although some RNN models showed little performance gains over the SVM baselines only with doc2vec features, they were even worse than the CRF model with the same features.

the RCNN models connecting the results of CNNs to the RNNs contributed to performance improvements not only from the baselines, but also from the CNN models.

contrasting

train_3716

While the CNN model tended to have a higher coverage in segmentation than the others, the LRCN model produced more precise decisions to recognize the boundaries on the strength of the consideration of conversational coherences in dialogue history sequences.

the segmentation performances even with the best models were still very limited especially for inter-categorical transitions.

contrasting

train_3717

Secondly, the use of external knowledge could be a key to success in dialogue topic tracking, as proved in the previous studies.

this work only takes internal dialogue information into account for making decisions.

contrasting

train_3718

Some of these studies also demonstrated that using L1specific properties, such as the error patterns of speakers of a given L1 or L1-induced paraphrases, improves the performance of automatic error correction in non-native writing.

neither of the approaches has constructed a semantic model from L1 data and systematically studied the effects of its transfer onto L2.

contrasting

train_3719

The results of these studies suggest that L1 is a valuable source of information in EDC.

all these works use isolated translational equivalents and focus on error correction only.

contrasting

train_3720

Adding the L1 lexicosemantic features has only a minor effect on accuracy and precision, and a more pronounced effect on recall.

when we test the system on one particular related L1 (Table 8) we observe the opposite effect: with the exception of ES subj data, precision and accuracy improve, suggesting that the error detection system using L1-induced information identifies errors more precisely.

contrasting

train_3721

For distractors considered "Plausible" by both experts, 63.6% were selected by the learners.

for those considered "Obviously wrong" by both experts, only 11.8% attracted any learner.

contrasting

train_3722

Limsopatham and Collier (2015a) leveraged translations between the informal language used in social media and the formal language used in the description of medical concepts in an ontology.

we argue that effective concept normalisation requires a system to take into account the semantics of social media messages and medical concepts.

contrasting

train_3723

We observe that for TwADR-S and TwADR-L datasets, which are smaller datasets (dataset size of 201 and 1,436, respectively), a better performance can be achieved if the model is not allowed to update the embeddings of the input phrases.

for the AskAPatient dataset (dataset size of 8,662), allowing the model to update the embeddings results in a significantly (paired t-test, p < 0.05) better performance.

contrasting

train_3724

We neglected such time-related phrases when defining C 0 .

the frequencies of the listed n-grams in the business group are much higher than those in the individual group.

contrasting

train_3725

For example, English POS taggers benefit from carefully designed word spelling features; orthographic features and external resources such as gazetteers are widely used in NER.

such task-specific knowledge is costly to develop (Ma and Xia, 2014), making sequence labeling models difficult to adapt to new tasks or new domains.

contrasting

train_3726

Several RNN-based neural network models have been proposed to solve sequence labeling tasks like speech recognition (Graves et al., 2013), POS tagging and NER (Chiu and Nichols, 2015;Hu et al., 2016), achieving competitive performance against traditional models.

even systems that have utilized distributed representations as inputs have used these to augment, rather than replace, hand-crafted features (e.g.

contrasting

train_3727

For many sequence labeling tasks it is beneficial to have access to both past (left) and future (right) contexts.

the LSTM's hidden state h t takes information only from past, knowing nothing about the future.

contrasting

train_3728

The RNNLM could be used in a KNN-style approach, where it associates each example response with its individual topic vector, using L(q) as a distance metric.

this is computationally infeasible since computing L(q) is significantly more expensive than cosine distance and the previously mentioned scalability would be lost.

contrasting

train_3729

SC-EX2, which relates to similar issues, is often confused with it.

sC-EX3 is rarely confused with sC-EX1 as it is about non-personal events on a larger scale.

contrasting

train_3730

A limitation of both the standard and proposed approach is that if a new question is created by the test-makers, then it will be necessary to collect example responses before it can be widely deployed.

since the system can be trained on ASR transcriptions, the example responses do not need to be hand-transcribed.

contrasting

train_3731

This makes automatic detection of sarcasm an important problem.

it has been quite difficult to solve such a problem with traditional NLP tools and techniques.

contrasting

train_3732

(2016) for modeling sarcasm understandability of readers.

as far as we know, these features are being introduced in NLP tasks like textual sarcasm detection for the first time.

contrasting

train_3733

(2013) conducted a corpus analysis and found certain similarity between Penn Discourse TreeBank relations (Prasad et al., 2008) and argumentation schemes (Walton et al., 2008).

they did not discuss how such similarity could be applied to argument mining.

contrasting

train_3734

Argument words (e.g., 'believe', 'opinion') signal the argumentative content and are commonly used across different topics.

domain words are specific terminologies commonly used within the topic (e.g., 'art', 'education').

contrasting

train_3735

If you know what labels you want, active learning can reduce the number of labeled documents needed.

establishing the label set remains difficult.

contrasting

train_3736

In LA and LR, the preference function directly chooses a document and directs the user to it.

u TA d and u TR d are topic dependent.

contrasting

train_3737

It is designed to detect error patterns from a fixed window of 7 words, which is large enough to not require the use of more advanced composition functions.

the performance of the bidirectional recurrent network (Bi-RNN) is somewhat lower, especially on the test set.

contrasting

train_3738

It is worth noting that the full Bi-LSTM has been trained on more data than the other CoNLL contestants.

as the shared task systems were not restricted to the NUCLE training set, all the submissions also used differing amounts of training data from various sources.

contrasting

train_3739

Through the availability of large annotated resources, such as PropBank (Palmer et al., 2005), statistical models based on such features achieve high accuracy.

results often fall short when the input to be labeled involves instances of linguistic phenomena that are relevant for the labeling decision but appear infrequently at training time.

contrasting

train_3740

Analogously to how word embeddings are computed, the simplest way to embed paths would be to represent each sequence as a one-hot vector.

this is suboptimal for two reasons: Firstly, we expect only a subset of dependency paths to be attested frequently in our data and therefore many paths will be too sparse to learn reliable embeddings for them.

contrasting

train_3741

The prior studies attempt to predict how strongly users are currently engaged in dialogues with systems.

the goal of this study is to predict how strongly users will be engaged with intelligent assistants in the future.

contrasting

train_3742

chat Voice Assist can give greetings to, have conversations with, and play games with users (e.g., V2 and V6).

to device operations for accomplishing certain tasks, these functions are offered for fun or for facilitating smooth communication.

contrasting

train_3743

In the trigger classification stage, some previous approaches (Grishman et al., 2005;Ji and Grishman, 2008;Liao and Grishman, 2010;Huang and Riloff, 2012) use patterns to decide the types of event triggers.

pattern-based approaches suffer from low recall since real world events usually have a large variety of representations.

contrasting

train_3744

DMCNN is mainly used to model contextual features.

dMCNN still does not consider argument-argument interactions.

contrasting

train_3745

We can see that RBPB with both plain feature and pattern feature can gain Table 3: The effect (F 1 value) of pattern feature much better performance than with two kinds of features alone.

our approach is just a pipeline approach which suffers from error propagation and the argument performance may not affect the trigger too much.

contrasting

train_3746

We find that JET depends too much on event patterns for event type priori and JET considers each candidate argument separately.

patterns cannot cover all events and the relationship between candidate arguments may help when identifying arguments.

contrasting

train_3747

Therefore, we develop an event type classifier to assign the event type, using both pattern matching information and other features, which gives our system the capability to deal with failed match cases when using patterns alone.

we train a maximum entropy classifier to predict the relationship between candidate arguments.

contrasting

train_3748

In summary, by using the event type classifier and the regularization method, we have achieved a good performance in which the trigger classification is comparable to state-of-theart methods, and the argument identification & classification performance is significantly better than state-of-the-art methods.

we only use sentence-level features and our method is a pipelined approach.

contrasting

train_3749

With self-training, we successfully exploit unlabeled data to improve over ENT by 15% F-score on the newswire domain, and 13% F-score on clinical data.

our active learning experiments demonstrate that we can match (and even beat) ENT using only 6.6% of the training data in the clinical domain, and only 5.8% of the training data in the newswire domain.

contrasting

train_3750

On RTE-2 their method achieves accuracy improvements of upto 10%.

hickl and Bensley (2007) achieves only a 1% accuracy improvement on RTE-3 using the same method, suggesting that it is not always as beneficial.

contrasting

train_3751

A token in one text passage that matches exactly, characterfor-character, with a token in another text passage would be considered a term match by this simple term matcher.

these term matchers could be more sophisticated and match pairs of terms that are synonyms, or paraphrases, or Table 3.

contrasting

train_3752

Our self-training system substantially improved over ENT, achieving an F-score gain of 15% on newswire and 13% on clinical, using only additional unlabeled data.

our active learning experiments demonstrated that we could match (and even beat) the baseline ENT system with only 6.6% of the training data in the clinical domain, and only 5.8% of the training data in the newswire domain.

contrasting

train_3753

Existing multilingual topic models consider only document-level alignments.

most documents are hierarchically structured, i.e., a document comprises segments (e.g., sections and paragraphs) that can be aligned across languages.

contrasting

train_3754

BiSTM leverages existing segments from a given segmentation.

a segmentation is not always given, and a given segmentation might not be optimal for statistical modeling.

contrasting

train_3755

This seems to be reasonable, because some original sections include multiple topics.

tables 2 and 3 show that inferred boundaries do not work better than section boundaries.

contrasting

train_3756

Orthogonal to the formal semantics of DCS, distributional vector representations are useful in capturing lexical semantics of words (Turney and Pantel, 2010;Levy et al., 2015), and progress is made in combining the word vectors to form meanings of phrases/sentences (Mitchell and Lapata, 2010; Baroni and Zamparelli, 2010;Grefenstette and Sadrzadeh, 2011;Socher et al., 2012;Paperno et al., 2014;Hashimoto et al., 2014).

less effort is devoted to finding a link between vector-based compositions and the composition operations in any formal semantics.

contrasting

train_3757

It suggests regularities in the semantic space, especially because orthogonal matrices preserve cosine similarity -if M N is orthogonal, two words x, y and their projections π N (x), π N (y) will have the same similarity measure, which is semantically reasonable.

matrices trained by vecUD are only orthogonal for three UD relations, namely conj, dep and appos.

contrasting

train_3758

(2014) and the entanglement model in Kartsaklis and Sadrzadeh (2014).

these models do not show particular advantages on other datasets.

contrasting

train_3759

Results As shown in Table 7, vecDCS scores better than the N-gram model and demonstrates promising performance.

to our surprise, "no matrix" shows an even better result which is the new state-of-the-art.

contrasting

train_3760

Logic and Distributional Semantics Logic is necessary for implementing the functional aspects of meaning and organizing knowledge in a structured and unambiguous way.

distributional semantics provides an elegant methodology for assessing semantic similarity and is well suited for learning from data.

contrasting

train_3761

(2014), in exploring vector calculations that realize logic operations.

the previous works did not specify how to integrate contextual distributional information, which is necessary for calculating semantic similarity.

contrasting

train_3762

word embedding) in the answer directly, and then we use RNN to model the attentive word sequence.

this model attends a sentence word by word which may ignore the relation between words.

contrasting

train_3763

In the example given above, the verb corresponds quite closely to the desired target relation.

in the wild, we encounter a multitude of different ways of expressing the same kind of relationship.

contrasting

train_3764

(2014) proposed a deep convolutional neural network with softmax classification, extracting lexical and sentence level features.

these approaches still depend on additional features from lexical resources and NLP toolkits.

contrasting

train_3765

In the training phase, we could probably find that every two entities connected by this relation are also connected by the path editor/film −1 , and hence assign an extremely high weight to it.

5 in the testing phase, for any entity pair (x, y) such that (y, editor/film, x) has not been encoded, we might not even find that path and hence could always make a negative prediction.

contrasting

train_3766

Website privacy policies are often ignored by Internet users, because these documents tend to be long and difficult to understand.

the significance of privacy policies greatly exceeds the attention paid to them: these documents are binding legal agreements between website operators and their users, and their opaqueness is a challenge not only to Internet users but also to policy regulators.

contrasting

train_3767

Legal text is a familiar domain for natural language processing, and the legal community has demonstrated some reciprocal interest (Mahler, 2015).

the scale of the problem and its significance-i.e., to virtually any Internet user, as well as to website operators and policy regulators-distinguishes it and provides immense motivation (Sadeh et al., 2013).

contrasting

train_3768

Other approaches leverage topic mod-eling (Chundi and Subramaniam, 2014;Stamey and Rossi, 2009) or sequence alignment techniques to analyze privacy policies or identify similar policy sections and paragraphs.

the complexity and vagueness of privacy policies makes it difficult to automatically extract complex data practices from privacy policies without substantial gold standard data.

contrasting

train_3769

The search is based on a so-called "floating parser" (Pasupat and Liang, 2015), a modification of a standard chart-parser, which is able to guide the search based on the similarity features.

our approach does not search among the derivations for the one that maximizes a match with the NL, but instead directly tries to predict a decision sequence that can be mapped to the LF.

contrasting

train_3770

number of rules) is larger than that of LF or CF.

dS is unique in allowing us to easily validate grammatical constraints.

contrasting

train_3771

gives small probability to ungrammatical choices, a property not shared by DSP-CL.

a more complete understanding of the difference will need more investigation.

contrasting

train_3772

Word embeddings-distributed representations of words-in deep learning are beneficial for many tasks in NLP.

different embedding sets vary greatly in quality and characteristics of the captured information.

contrasting

train_3773

Each embedding set contains information about only a part of the overall vocabulary.

it can predict what the remaining part should look like by comparing words it knows with the information other embedding sets provide about these words.

contrasting

train_3774

(i) Not all individual embedding sets are beneficial in this task; e.g., HLBL embeddings make FLORS perform worse in 11 out of 12 cases.

(ii) in most cases, embeddings improve system performance, which is consistent with prior work on using embeddings for this type of task (Xiao and Guo, 2013;Yang and Eisenstein, 2014;Tsuboi, 2014).

contrasting

train_3775

As a concrete example we collect live text data and corresponding news reports for football (called soccer more often in the United States) games and conduct our study thereby.

our methods and discussions made in this paper can be trivially adapted to other types of sports games as well.

contrasting

train_3776

This is used for traditional summarization problems (Kulesza and Taskar, 2011).

the problem for constructing sports news from live broadcast script is rather different.

contrasting

train_3777

The comparison between RF and RF+DPP shows the effectiveness of our sentence selection strategy.

the increase is still limited 12 .

contrasting

train_3778

All the sentences are describing a key scoring event.

none of them were selected to construct the news because our LTR model assigns low scores for these short sentences.

contrasting

train_3779

In our experiments, methods with higher ROUGE scores can indeed achieve better coverage of important units such as events, as shown in pyramid scores in Table 2.

we can also observe from Table 5 that automatic metrics currently cannot reflect readability factors very well.

contrasting

train_3780

(2010) introduced a joint rule selection model for hierarchical phrase-based translation, which also approximated the rule selection problem by a binary classification problem like our approach.

these two models adopted linear classifiers similar to those used in the MERS model , which suffers more from the data sparsity problem compared to the CSRS model.

contrasting

train_3781

In traditional linear models, combining scores can be implemented by including low-order features.

for neural models, this is not that straightforward because of nonlinearity.

contrasting

train_3782

The neural rerankers are capable of capturing global syntax features across the tree.

the most non-local neural parser with LSTM cannot exploit global features.

contrasting

train_3783

These methods aim to learn common feature representations for different languages.

most of the current researches only focus on bilingual word embedding.

contrasting

train_3784

grSemi-CRFs use much fewer input features and most of them are task-insensitive 13 .

grSemi-CRFs achieve almost the same performance, sometimes even better.

contrasting

train_3785

For text chunking, grSemi-CRF outperforms all reported supervised models, except JESS-CM (Suzuki and Isozaki, 2008), a semi-supervised model using giga-word scale unlabeled data in training 14 .

the performance of our grSemi-CRF (95.01%) is very close to that of JESS-CM (95.15%).

contrasting

train_3786

Summarization of event descriptions can be improved by selecting causally motivated sentences.

causality is frequently expressed implicitly, which requires world knowledge and inference.

contrasting

train_3787

The explosion caused people to evacuate the building.

the words can not be substituted in the following sentence: The baker made a cake.

contrasting

train_3788

Similar to the FrameNet features, we split the example into three sections.

we also consider the dependency parse of the data.

contrasting

train_3789

VerbNet VerbNet is a resource devoted to storing information for verbs (Kipper et al, 2000).

to WordNet, VerbNet provides a more fine-grained description of events while focusing less on polysemy.

contrasting

train_3790

Modeling relation paths has offered significant gains in embedding models for knowledge base (KB) completion.

enumerating paths between two entities is very expensive, and existing approaches typically resort to approximation with a sampled subset.

contrasting

train_3791

Therefore prior work selected a subset of paths through sampling and pruning (Neelakantan et al., 2015;Lin et al., 2015).

the properties of the BILINEAR composition function for path representation enable us to incrementally build the sums of all path representations exactly, using dynamic programming.

contrasting

train_3792

In either case, the accuracy difference between ALL-PATHS and PRUNED-PATHS is small, and ALL-PATHS mainly gains in efficiency.

when nodes are modeled, the compositional learning approach gains in accuracy as well, especially when text is jointly embedded.

contrasting

train_3793

For the DNN models, we add the penalty term λ θ 2 to the losses, where λ is the regularization coefficient and θ contains all other parameters.

for the bilinear models we regularize the relation matrices M R toward the identity matrix instead of all zeroes, adding the following to the loss: where I r is the r × r identity matrix, the summation is performed over all relations R, and θ represents all other parameters.

contrasting

train_3794

Figure 6 shows how the models perform as the number of utterances per example varies.

when the search space is small (fewer number of utterances), Model B outperforms or is competitive with Model C. as the search space increases (tighter computational constraints), Model C does increasingly better.

contrasting

train_3795

Even though Model A is unable to learn anything with beam size < 240.

for beam sizes larger than 240, Model A attains 100% accuracy.

contrasting

train_3796

Recall that projecting from Model A to C creates a more computationally tractable model at the cost of expressivity.

this is because Model C used a linear model.

contrasting

train_3797

Tree-structured neural networks exploit valuable syntactic parse information as they interpret the meanings of sentences.

they suffer from two key technical problems that make them slow and unwieldy for large-scale NLP tasks: they usually operate on parsed sentences and they do not directly support batched computation.

contrasting

train_3798

Our results show that a model that uses tree-structured composition fully (SPINN) outperforms one which uses it only partially (tree-based CNN), which in turn outperforms one which does not use it at all (RNN).

the full SPINN performed moderately well at reproducing the Stanford Parser's parses of the SNLI data at a transition-by-transition level, with 92.4% accuracy at test time.3 its transi-3Note that this is scoring the model against automatic tion prediction errors are fairly evenly distributed across sentences, and most sentences were assigned partially invalid transition sequences that either left a few words out of the final representation or incorporated a few padding tokens into the final representation.

contrasting

train_3799

For SG model, the max selectivity of the model neurons is only just above 0.7.

the context-based distributed models showed strong selective activation towards country names in Indonesian.

contrasting