Design and Development of the Sentence-based Collocation Recommender with Error Detection for Academic Writing

Abstract

Appropriate collocations in writing of research papers in English can make the context smoother and expression of ideas more precise. In consequence, it is easier for the reader to understand and the purpose of sharing the outcome of research is accomplished. However, for non-native English speakers, the choice and use of collocations is very difficult. For this reason, this study is intended to refer to a large amount of high-quality academic literature to establish a collocation corpus and adopt natural language processing techniques and statistical methods to develop a collocation recommendation system. The system will allow users to enter sentences, automatically detect the locations and types of collocations and recommend synonymous collocations in accordance with the semantics and frequency of use. The fitness of collocations in the system for beginning sentences achieves 73.1%. Writers of academic papers can use the system to select appropriate collocations, reduce erroneous use of collocations and improve the quality of their papers.