Modelling context with neural networks for recommending idioms in essay writing

Although the proper use of idioms can enhance the conciseness and elegance of writing, actively using idioms in writing essays differs from understanding them in their contexts. Building an AI system for idiom recommendation is also challenging as idioms and their surrounding contexts rarely share words or characters, and many idioms even come from legends or historic stories. In this study, we propose a neural network-based approach to address the novel task of recommending idioms in essay writing. We encode semantic representations of variable-sized contexts by considering global and local contexts and modelling them with different schemes. The goal is to calculate the similarity between the given context and candidate idioms. Experimental results demonstrate the effectiveness of the proposed approach compared with other state-of-the-art methods. (c) 2017 Elsevier B.V. All rights reserved.