In theory, grammar checkers should be great aids to language learning. In practice, however, teachers and students have found them to be disappointing computer-assisted language learning aids. There are a number of reasons for this disappointment: the computational design task is extremely difficult, many vendors have cut comers in the rush to market, the technology has not been well-integrated with the realities of classroom management, and the appropriate role of grammar checkers has been misunderstood. After a brief discussion of the former problems, this paper examines the last problem. Researchers, teachers, and students have been led to expect that computers will perform like humans in the language learning process. But whereas human teachers are good at error analysis, grammar-checkers are best-suited to competency analysis: human teachers are best at correcting student errors, while computers are best at evaluating students' competence. We describe our own efforts at constructing error-analyzing grammar checkers for English, Chinese, Russian, Japanese, German, Hindi, and a range of other languages, and illustrate their use for both error and competency analyses. While conceding that error-analyzing grammar checkers can have some value for some students, we will argue that their primary role should be that of evaluating and modeling students' progress in language learning.
展开▼