An Approach toward Automatic Error Detection in Learners’ English Writing Based on the Source Language

Automatic error detection systems for English writing have been improved since they were first introduced and are being applied to foreign language learning. However, these systems mainly focus on local errors, such as grammatical aspects in the target language and ignore the meaning intended in the source language. As a result, teachers must spend an inordinate amount of time to detect global errors. In this paper, we propose an approach to an automatic error detection system to solve this problem. In order to determine whether the structure of an English sentence is in error or not, criteria for error determination are first needed. Our approach is based on the idea that criteria for error determination are created by the correspondence relation between Japanese and English using sentence patterns. In order to evaluate our approach, by way of illustration, four sentence patterns were selected from the authors’ original six sentence patterns. Automatic error detection using these four sentence patterns was carried out on 100 Japanese sentences with subjects and their corresponding English sentences. As a result, we concluded that, using the sentence patterns in the source language, automatic error detection is effective when based on our criteria for error determination.