CONVERSATION REBUILDING - FROM THE FOREIGN-LANGUAGE CLASSROOM TO IMPLEMENTATION IN AN INTELLIGENT TUTORING SYSTEM

This paper shows how an innovative ''communicative'' technique in teac hing foreign languages-Conversation Rebuilding (CR)-readily lends itse lf to implementation in an Intelligent Tutoring System (ITS). Classroo m language teachers using CR get students to formulate acceptable utte rances in a foreign idiom by starting from rough approximations (using words the students know) and gradually zeroing in on the utterance wh ich a native speaker of that idiom might produce in a similar setting. The ITS presented here helps students do the ''zeroing in'' optimally . It lets them express themselves temporarily in an ''interlingua'' (i .e., in their own kind of French or English or whatever they are study ing), as long as they make something of their communicative intent cle ar, that is, as long as the System can find a semantic starting point on which to build. The ITS then prods the students to express themselv es more intelligibly, starting from the ''key'' elements (determined b y a heuristic based on how expert classroom teachers proceed) and taki ng into consideration the students' past successful or unsuccessful at tempts at communication. To simplify system design and programming, ho wever, conversations are ''constrained'': students playact characters in set dialogs and aim at coming up with what the characters actually say (not what they could possibly say). While most Intelligent Compute r Assisted Language Learning (ICALL) focuses the attention of students on norms to acquire, the ICALL implementation of CR presented in this paper focuses the attention of students on saying something-indeed, a lmost anything-to keep the conversation going and get some kind of mea ning across to the other party. It sees successful language acquisitio n primarily as the association of forms with intent, not simply as the conditioning of appropriate reflexes or the elaboration/recall of con ceptualized rules (which are the by-products of successful communicati on). Thus, in espousing this hard-line communicative approach, the pre sent paper makes a first, non-trivial point: ICALL researchers might u sefully begin by investigating what the more able teachers are doing i n the classroom, rather than by building elaborate computer simulation s of out-dated practices, as happens all too often. The paper then goe s on to describe the architecture of a prototype ITS based on CR-one t hat the authors have actually implemented and tested-for the acquisiti on of English as a foreign language. A sample learning session is tran scribed to illustrate the man-machine interaction. Concluding remarks show how the present-day limits of ICALL (and Artificial Intelligence in general) can be partially circumvented by the strategy implemented in the program, i.e. by making the students feel they are creatively p iloting an interaction rather than being tested by an unimaginative ma chine. (C) 1997 Elsevier Science Ltd. All rights reserved.