Discourse Marker Augmented Network with Reinforcement Learning for Natural Language Inference

Abstract

Natural Language Inference (NLI), also known as Recognizing Textual Entailment (RTE), is one of the most important problems in natural language processing. It requires to infer the logical relationship between two given sentences. While current approaches mostly focus on the interaction architectures of the sentences, in this paper, we propose to transfer knowledge from some important discourse markers to augment the quality of the NLI model. We observe that people usually use some discourse markers such as "so" or "but" to represent the logical relationship between two sentences. These words potentially have deep connections with the meanings of the sentences, thus can be utilized to help improve the representations of them. Moreover, we use reinforcement learning to optimize a new objective function with a reward defined by the property of the NLI datasets to make full use of the labels information. Experiments show that our method achieves the state-of-the-art performance on several large-scale datasets.

arXiv commentAccepted in ACL 2018

References in corpus (10)

• Distributed Representations of Words and Phrases and their Compositionality
• ADADELTA: An Adaptive Learning Rate Method
• Enhanced LSTM for Natural Language Inference
• Skip-Thought Vectors
• Aligning Books and Movies: Towards Story-like Visual Explanations by Watching Movies and Reading Books
• Neural Natural Language Inference Models Enhanced with External Knowledge
• Discourse-Based Objectives for Fast Unsupervised Sentence Representation Learning
• MEMEN: Multi-layer Embedding with Memory Networks for Machine Comprehension
• DisSent: Sentence Representation Learning from Explicit Discourse Relations
• Keyword-based Query Comprehending via Multiple Optimized-Demand Augmentation