Learning Natural Language Inference with LSTM

Proceedings of NAACL-HLT 2016, pages 1442–1451, San Diego, California, June 12-17, 2016. c⃝2016 Association for Computational Linguistics Learning Natural Language Inference with LSTM Shuohang Wang School of Information Systems Singapore Management University shwang.2014@phdis.smu.edu.sg Jing Jiang School of Information Systems Singapore Management University jingjiang@smu.edu.sg Abstract Natural language inference (NLI) is a funda- mentally important task in natural language processing that has many applications. The recently released Stanford Natural Language Inference (SNLI) corpus has made it possi- ble to develop and evaluate learning-centered methods such as deep neural networks for nat- ural language inference (NLI). In this paper, we propose a special long short-term mem- ory (LSTM) architecture for NLI. Our model builds on top of a recently proposed neural at- tention model for NLI but is based on a sig- nificantly different idea. Instead of deriving sentence embeddings for the premise and the hypothesis to be used for classification, our so- lution 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. 1 Introduction Natural language inference (NLI) is the problem of determining whether from a premise sentence P one can infer another hypothesis sentence H (MacCart- ney, 2009). NLI is a fundamentally important prob- lem that has applications in many tasks including question answering, semantic search and automatic text summarization. There has been much inter- est in NLI in the past decade, especially surround- ing the PASCAL Recognizing Textual Entailment (RTE) Challenge (Dagan et al., 2005). Existing so- lutions to NLI range from shallow approaches based on lexical similarities (Glickman et al., 2005) to ad- vanced methods that consider syntax (Mehdad et al., 2009), perform explicit sentence alignment (Mac- Cartney et al., 2008) or use formal logic (Clark and Harrison, 2009). Recently, Bowman et al. (2015) released the Stan- ford Natural Language Inference (SNLI) corpus for the purpose of encouraging more learning-centered approaches to NLI. This corpus contains around 570K sentence pairs with three labels: entailment, contradiction and neutral. The size of the corpus makes it now feasible to train deep neural network models, which typically require a large amount of training data. Bowman et al. (2015) tested a straight- forward architecture of deep neural networks for NLI. In their architecture, the premise and the hy- pothesis are each represented by a sentence embed- ding vector. The two vectors are then fed into a multi-layer neural network to train a classifier. Bow- man et al. (2015) achieved an accuracy of 77.6% when long short-term memory (LSTM) networks were used to obtain the sentence embeddings. A more recent work by Rockt¨aschel et al. (2016) improved the performance by applying a neural at- tention model. While their basic architecture is still based on sentence embeddings for …

Content truncated to 500 words