Automated Error Detection in English Examination Essays Written by Russian Students

The recent development of computer-assisted language learning has brought about the need for automatic feedback and evaluation of the learners’ essays, which implies automatic error detection and correction. Certain types of second language errors (e.g. article and preposition errors) have received significant attention from researchers, whereas others (e.g. verb errors) have largely remained understudied so far. In the present work, I propose a system of error detection and correction for essays written by Russian students learning English as foreign language. I compare several major approaches to automated error annotation and combine the best models based on unannotated native language and annotated second language data. The project aims to design a system that could be integrated with essay upload functionality and would generate annotations for several types of errors yielding high precision. Automatic error annotation preprocessing reduces the time cost of manual annotation and gives students and teachers preliminary instant evaluation of the essay. The system is intended as a part of the automatic feedback system which is currently being developed.