The ability to reason with natural language is a fundamental prerequisite for many NLP tasks such as information extraction, machine translation and question answering. To quantify this ability, systems are commonly tested whether they can recognize textual entailment, i.e., whether one sentence can be inferred from another one. However, in most NLP applications only single source sentences instead of sentence pairs are available. Hence, we propose a new task that measures how well a model can generate an entailed sentence from a source sentence. We take entailment-pairs of the Stanford Natural Language Inference corpus and train an LSTM with attention. On a manually annotated test set we found that 82% of generated sentences are correct, an improvement of 10.3% over an LSTM baseline. A qualitative analysis shows that this model is not only capable of shortening input sentences, but also inferring new statements via paraphrasing and phrase entailment. We then apply this model recursively to input-output pairs, thereby generating natural language inference chains that can be used to automatically construct an entailment graph from source sentences. Finally, by swapping source and target sentences we can also train a model that given an input sentence invents additional information to generate a new sentence.
We seek to address the lack of labeled data (and high cost of annotation) for textual entailment in some domains. To that end, we first create (for experimental purposes) an entailment dataset for the clinical domain, and a highly competitive supervised entailment system, ENT, that is effective (out of the box) on two domains. We then explore self-training and active learning strategies to address the lack of labeled data. 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. On the other hand, 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.
NLP tasks differ in the semantic information they require, and at this time no single se- mantic representation fulfills all requirements. Logic-based representations characterize sentence structure, but do not capture the graded aspect of meaning. Distributional models give graded similarity ratings for words and phrases, but do not capture sentence structure in the same detail as logic-based approaches. So it has been argued that the two are complementary. We adopt a hybrid approach that combines logic-based and distributional semantics through probabilistic logic inference in Markov Logic Networks (MLNs). In this paper, we focus on the three components of a practical system integrating logical and distributional models: 1) Parsing and task representation is the logic-based part where input problems are represented in probabilistic logic. This is quite different from representing them in standard first-order logic. 2) For knowledge base construction we form weighted inference rules. We integrate and compare distributional information with other sources, notably WordNet and an existing paraphrase collection. In particular, we use our system to evaluate distributional lexical entailment approaches. We use a variant of Robinson resolution to determine the necessary inference rules. More sources can easily be added by mapping them to logical rules; our system learns a resource-specific weight that corrects for scaling differences between resources. 3) In discussing probabilistic inference, we show how to solve the inference problems efficiently. To evaluate our approach, we use the task of textual entailment (RTE), which can utilize the strengths of both logic-based and distributional representations. In particular we focus on the SICK dataset, where we achieve state-of-the-art results.
Many important NLP problems can be posed as dual-sequence or sequence-to-sequence modeling tasks. Recent advances in building end-to-end neural architectures have been highly successful in solving such tasks. In this work we propose a new architecture for dual-sequence modeling that is based on associative memory. We derive AM-RNNs, a recurrent associative memory (AM) which augments generic recurrent neural networks (RNN). This architecture is extended to the Dual AM-RNN which operates on two AMs at once. Our models achieve very competitive results on textual entailment. A qualitative analysis demonstrates that long range dependencies between source and target-sequence can be bridged effectively using Dual AM-RNNs. However, an initial experiment on auto-encoding reveals that these benefits are not exploited by the system when learning to solve sequence-to-sequence tasks which indicates that additional supervision or regularization is needed.
This work studies comparatively two typical sentence pair classification tasks: textual entailment (TE) and answer selection (AS), observing that phrase alignments of different intensities contribute differently in these tasks. We address the problems of identifying phrase alignments of flexible granularity and pooling alignments of different intensities for these tasks. Examples for flexible granularity are alignments between two single words, between a single word and a phrase and between a short phrase and a long phrase. By intensity we roughly mean the degree of match, it ranges from identity over surface-form co-occurrence, rephrasing and other semantic relatedness to unrelated words as in lots of parenthesis text. Prior work (i) has limitations in phrase generation and representation, or (ii) conducts alignment at word and phrase levels by handcrafted features or (iii) utilizes a single attention mechanism over alignment intensities without considering the characteristics of specific tasks, which limits the system's effectiveness across tasks. We propose an architecture based on Gated Recurrent Unit that supports (i) representation learning of phrases of arbitrary granularity and (ii) task-specific focusing of phrase alignments between two sentences by attention pooling. Experimental results on TE and AS match our observation and are state-of-the-art.
We consider the problem of Recognizing Textual Entailment within an Information Retrieval context, where we must simultaneously determine the relevancy as well as degree of entailment for individual pieces of evidence to determine a yes/no answer to a binary natural language question. We compare several variants of neural networks for sentence embeddings in a setting of decision-making based on evidence of varying relevance. We propose a basic model to integrate evidence for entailment, show that joint training of the sentence embeddings to model relevance and entailment is feasible even with no explicit per-evidence supervision, and show the importance of evaluating strong baselines. We also demonstrate the benefit of carrying over text comprehension model trained on an unrelated task for our small datasets. Our research is motivated primarily by a new open dataset we introduce, consisting of binary questions and news-based evidence snippets. We also apply the proposed relevance-entailment model on a similar task of ranking multiple-choice test answers, evaluating it on a preliminary dataset of school test questions as well as the standard MCTest dataset, where we improve the neural model state-of-art.
Introducing attentional mechanism in neural network is a powerful concept, and has achieved impressive results in many natural language processing tasks. However, most of the existing models impose attentional distribution on a flat topology, namely the entire input representation sequence. Clearly, any well-formed sentence has its accompanying syntactic tree structure, which is a much rich topology. Applying attention to such topology not only exploits the underlying syntax, but also makes attention more interpretable. In this paper, we explore this direction in the context of natural language inference. The results demonstrate its efficacy. We also perform extensive qualitative analysis, deriving insights and intuitions of why and how our model works.
Parkinson's disease is the second most common neurodegenerative disease, affecting more than 1.2 million people in Europe. Medications are available for the management of its symptoms, but the exact cause of the disease is unknown and there is currently no cure on the market. To better understand the relations between new findings and current medical knowledge, we need tools able to analyse published medical papers based on natural language processing and tools capable to identify various relationships of new findings with the current medical knowledge. Our work aims to fill the above technological gap. To identify conflicting information in medical documents, we enact textual entailment technology. To encapsulate existing medical knowledge, we rely on ontologies. To connect the formal axioms in ontologies with natural text in medical articles, we exploit ontology verbalisation techniques. To assess the level of disagreement between human agents with respect to a medical issue, we rely on fuzzy aggregation. To harmonize this disagreement, we design mediation protocols within a multi-agent framework.
In this paper, we propose a new deep learning approach, called neural association model (NAM), for probabilistic reasoning in artificial intelligence. We propose to use neural networks to model association between any two events in a domain. Neural networks take one event as input and compute a conditional probability of the other event to model how likely these two events are to be associated. The actual meaning of the conditional probabilities varies between applications and depends on how the models are trained. In this work, as two case studies, we have investigated two NAM structures, namely deep neural networks (DNN) and relation-modulated neural nets (RMNN), on several probabilistic reasoning tasks in AI, including recognizing textual entailment, triple classification in multi-relational knowledge bases and commonsense reasoning. Experimental results on several popular datasets derived from WordNet, FreeBase and ConceptNet have all demonstrated that both DNNs and RMNNs perform equally well and they can significantly outperform the conventional methods available for these reasoning tasks. Moreover, compared with DNNs, RMNNs are superior in knowledge transfer, where a pre-trained model can be quickly extended to an unseen relation after observing only a few training samples. To further prove the effectiveness of the proposed models, in this work, we have applied NAMs to solving challenging Winograd Schema (WS) problems. Experiments conducted on a set of WS problems prove that the proposed models have the potential for commonsense reasoning.
In the context of the Competition on Legal Information Extraction/Entailment (COLIEE), we propose a method comprising the necessary steps for finding relevant documents to a legal question and deciding on textual entailment evidence to provide a correct answer. The proposed method is based on the combination of several lexical and morphological characteristics, to build a language model and a set of features for Machine Learning algorithms. We provide a detailed study on the proposed method performance and failure cases, indicating that it is competitive with state-of-the-art approaches on Legal Information Retrieval and Question Answering, while not needing extensive training data nor depending on expert produced knowledge. The proposed method achieved significant results in the competition, indicating a substantial level of adequacy for the tasks addressed.
In this paper we address the question of how to render sequence-level networks better at handling structured input. We propose a machine reading simulator which processes text incrementally from left to right and performs shallow reasoning with memory and attention. The reader extends the Long Short-Term Memory architecture with a memory network in place of a single memory cell. This enables adaptive memory usage during recurrence with neural attention, offering a way to weakly induce relations among tokens. The system is initially designed to process a single sequence but we also demonstrate how to integrate it with an encoder-decoder architecture. Experiments on language modeling, sentiment analysis, and natural language inference show that our model matches or outperforms the state of the art.
Evaluation of NLP methods requires testing against a previously vetted gold-standard test set and reporting standard metrics (accuracy/precision/recall/F1). The current assumption is that all items in a given test set are equal with regards to difficulty and discriminating power. We propose Item Response Theory (IRT) from psychometrics as an alternative means for gold-standard test-set generation and NLP system evaluation. IRT is able to describe characteristics of individual items - their difficulty and discriminating power - and can account for these characteristics in its estimation of human intelligence or ability for an NLP task. In this paper, we demonstrate IRT by generating a gold-standard test set for Recognizing Textual Entailment. By collecting a large number of human responses and fitting our IRT model, we show that our IRT model compares NLP systems with the performance in a human population and is able to provide more insight into system performance than standard evaluation metrics. We show that a high accuracy score does not always imply a high IRT score, which depends on the item characteristics and the response pattern.
We propose a simple neural architecture for natural language inference. Our approach uses attention to decompose the problem into subproblems that can be solved separately, thus making it trivially parallelizable. On the Stanford Natural Language Inference (SNLI) dataset, we obtain state-of-the-art results with almost an order of magnitude fewer parameters than previous work and without relying on any word-order information. Adding intra-sentence attention that takes a minimum amount of order into account yields further improvements.
Several tasks in argumentation mining and debating, question-answering, and natural language inference involve classifying a sequence in the context of another sequence (referred as bi-sequence classification). For several single sequence classification tasks, the current state-of-the-art approaches are based on recurrent and convolutional neural networks. On the other hand, for bi-sequence classification problems, there is not much understanding as to the best deep learning architecture. In this paper, we attempt to get an understanding of this category of problems by extensive empirical evaluation of 19 different deep learning architectures (specifically on different ways of handling context) for various problems originating in natural language processing like debating, textual entailment and question-answering. Following the empirical evaluation, we offer our insights and conclusions regarding the architectures we have considered. We also establish the first deep learning baselines for three argumentation mining tasks.
Machine comprehension of text is an important problem in natural language processing. A recently released dataset, the Stanford Question Answering Dataset (SQuAD), offers a large number of real questions and their answers created by humans through crowdsourcing. SQuAD provides a challenging testbed for evaluating machine comprehension algorithms, partly because compared with previous datasets, in SQuAD the answers do not come from a small set of candidate answers and they have variable lengths. We propose an end-to-end neural architecture for the task. The architecture is based on match-LSTM, a model we proposed previously for textual entailment, and Pointer Net, a sequence-to-sequence model proposed by Vinyals et al.(2015) to constrain the output tokens to be from the input sequences. We propose two ways of using Pointer Net for our task. Our experiments show that both of our two models substantially outperform the best results obtained by Rajpurkar et al.(2016) using logistic regression and manually crafted features.
The utilization of social media material in journalistic workflows is increasing, demanding automated methods for the identification of mis- and disinformation. Since textual contradiction across social media posts can be a signal of rumorousness, we seek to model how claims in Twitter posts are being textually contradicted. We identify two different contexts in which contradiction emerges: its broader form can be observed across independently posted tweets and its more specific form in threaded conversations. We define how the two scenarios differ in terms of central elements of argumentation: claims and conversation structure. We design and evaluate models for the two scenarios uniformly as 3-way Recognizing Textual Entailment tasks in order to represent claims and conversation structure implicitly in a generic inference model, while previous studies used explicit or no representation of these properties. To address noisy text, our classifiers use simple similarity features derived from the string and part-of-speech level. Corpus statistics reveal distribution differences for these features in contradictory as opposed to non-contradictory tweet relations, and the classifiers yield state of the art performance.
Multi-hop inference is necessary for machine learning systems to successfully solve tasks such as Recognising Textual Entailment and Machine Reading. In this work, we demonstrate the effectiveness of adaptive computation for learning the number of inference steps required for examples of different complexity and that learning the correct number of inference steps is difficult. We introduce the first model involving Adaptive Computation Time which provides a small performance benefit on top of a similar model without an adaptive component as well as enabling considerable insight into the reasoning process of the model.
Natural language inference (NLI) is a fundamentally important task in natural language processing that has many applications. The recently released Stanford Natural Language Inference (SNLI) corpus has made it possible to develop and evaluate learning-centered methods such as deep neural networks for natural language inference (NLI). In this paper, we propose a special long short-term memory (LSTM) architecture for NLI. Our model builds on top of a recently proposed neural attention model for NLI but is based on a significantly different idea. Instead of deriving sentence embeddings for the premise and the hypothesis to be used for classification, our solution uses a match-LSTM to perform word-by-word matching of the hypothesis with the premise. This LSTM is able to place more emphasis on important word-level matching results. In particular, we observe that this LSTM remembers important mismatches that are critical for predicting the contradiction or the neutral relationship label. On the SNLI corpus, our model achieves an accuracy of 86.1%, outperforming the state of the art.
Deep learning techniques are increasingly popular in the textual entailment task, overcoming the fragility of traditional discrete models with hard alignments and logics. In particular, the recently proposed attention models (Rockt\"aschel et al., 2015; Wang and Jiang, 2015) achieves state-of-the-art accuracy by computing soft word alignments between the premise and hypothesis sentences. However, there remains a major limitation: this line of work completely ignores syntax and recursion, which is helpful in many traditional efforts. We show that it is beneficial to extend the attention model to tree nodes between premise and hypothesis. More importantly, this subtree-level attention reveals information about entailment relation. We study the recursive composition of this subtree-level entailment relation, which can be viewed as a soft version of the Natural Logic framework (MacCartney and Manning, 2009). Experiments show that our structured attention and entailment composition model can correctly identify and infer entailment relations from the bottom up, and bring significant improvements in accuracy.
We present a memory augmented neural network for natural language understanding: Neural Semantic Encoders. NSE is equipped with a novel memory update rule and has a variable sized encoding memory that evolves over time and maintains the understanding of input sequences through read}, compose and write operations. NSE can also access multiple and shared memories. In this paper, we demonstrated the effectiveness and the flexibility of NSE on five different natural language tasks: natural language inference, question answering, sentence classification, document sentiment analysis and machine translation where NSE achieved state-of-the-art performance when evaluated on publically available benchmarks. For example, our shared-memory model showed an encouraging result on neural machine translation, improving an attention-based baseline by approximately 1.0 BLEU.