An automatic text-analysis project for EFL writing revision

Abstract

In order to leave more time for EFL teachers to work on higher-level re-writing tasks, we have launched a project to develop an automatic English grammar text-analysis program for Chinese students in Taiwan, Republic of China to help writing revision processes. This year-long project included extensive research work in several areas: error analysis, linguistic analysis of English, and natural language processing in computational linguistics. The initial effort was devoted to error analysis of 135 writing samples collected from our students mainly with engineering backgrounds. Most of the error types were then formulated into patterns to facilitate the on-line implementation of the grammar program. In addition, linguistic analyses in the areas of idiom collection/analyses, subcategorization of word categories (such as verbs and adverbs), and lexical feature analyses were performed. A small electronic dictionary with 2066 word stems and necessary features was also built, and a suffix processor with morphological rules to accommodate morpho-syntactic variants of each word stem was designed. A parser, equipped with phrase structure rules and error patterns, was then constructed. In addition, appropriate feedback messages were provided when the program located an error. Initial achievements indicated that the grammar program performed slightly better than commercial packages.