基于TAACO的文本连贯性与二语写作质量关系研究

A Study on Relationship between Text Coherence and Second Language Writing Quality Based on TAACO

摘要: 随着自然语言处理技术的发展，文本连贯性的测量与研究已经成为二语写作评价的焦点。本研究以英语专业八级考试的写作文本为语料，探究中国高水平英语学习者的文本连贯性与写作质量的关系。研究发现：1) 局部连贯性与整体连贯性和写作质量呈正相关，其中相关性最强的指标为Word2vec，而篇章连贯性与二语写作质量不具有相关性；2) 整体Word2vec、局部论元重叠、整体词元重叠等三项指标的联合可以解释写作成绩32.8%的方差。研究结论一定程度上可启示我国英语写作教学，也可为作文自动评分工具的开发提供参考。

Abstract: With the development of natural language processing technology, the measurement and study of text coherence has become the focus of second language writing evaluation. Taking TEM8 writings as the research corpus, this study explores the relationship between text coherence and writing quality among high-level Chinese EFL learners. The results show that: 1) There is a positive correlation between local coherence and global coherence and writing quality, of which Word2vec is the most correlative indicator, while there is no correlation between text coherence and second language writing quality; 2) The combination of global Word2vec, local argument overlap and global lemmas overlap can explain 32.8% of the variance of writing performance. The conclusion of the study can enlighten the teaching of English writing in China to some extent, and also provide a reference for the development of automatic writing scoring tools.