Applying computer models of phonological competence to C.A.L.L.

Abstract

The following paper attempts to show that a computer model of a native speaker's phonological competence, while generating significant improvements to the theory itself provides an adequate framework for intelligent Computer Assisted Language Learning. Allowing automatic error detection, self-adjusting correction and treatment of a large corpus of words, a program incorporating formalized linguistic knowledge does not only offer more simplicity, power and freedom than would more conventional language teaching programs; it also departs radically from a traditionally structuralist approach of C.A.L.L. based on “pattern practice” and conditioning rather than the acquisition of an explicit linguistic knowledge. We will successively discuss two formal rules of English phonology, their computer modelization and their insertion in an experimental teaching program.