Adaptive modelling of student diagnosis and material selection for on-line language learning

The use of interactive on-line environments in distance learning appears to be promising because they solve some of the difficulties teachers have maintaining control of individual student progress due to the large number of students present in such courses. However, in the case of language learning, work is still required to design a system that is as flexible and effective as an experienced teacher. In this paper I-PETER (Intelligent Personalised English Tutoring EnviRonment) is presented as an advance in this area. It has four domain models that represent linguistic and didactic knowledge. Its student linguistic knowledge model is richer than that typically used in language teaching: a student's command of English is evaluated by interpreting his/her performance on specific linguistic units in terms of three related criteria, rather than by a general linguistic competence ranking. This model enables error diagnosis to be undertaken using a Bayesian network, to reflect how teachers actually undertake the process in the classroom. The results of this diagnosis process enable a finer-grained control of material selection than is normally possible, giving rise to a course structure that is continuously adapted to individual student needs.