Empirical studies and intelligent language tutoring

The research described in this paper addresses how a language tutoring system tackles a practical problem in learning a language — negative transfer (mother tongue influence). The empirical studies we have undertaken indicate that negative transfer is the most obvious explanation for most of the errors committed by first-year students in their study of Chinese grammar at the University of Durham. The results of our empirical studies accord with the views of experts on both negative transfer and the Chinese language. A language tutoring prototype (the Chinese Tutor) has been designed on the basis of the results of the empirical studies. The Chinese Tutor contains five models: the expert model, the student model, the diagnoser, the tutor model and the interface module. The five models work effectively together in order to ensure a successful tutorial session. A number of potential users from the Department of East Asian Studies at the University of Durham have used the Chinese Tutor with positive results. In this paper, we will first address issues in relation to empirical studies: the significance of empirical studies and the results of the empirical studies. We will then move on to some of the important design issues, i.e., how the problem of negative transfer is tackled by the Chinese Tutor, which includes discussions on how the rules in the mixed grammar are used for detecting arbitrary transfer errors; how the types of error made by the student are inferred; and how an appropriate tutorial strategy is selected. Finally, a tutorial session run by one of the students who has used the Chinese Tutor is presented and what the Chinese Tutor can offer to the students will also be addressed.