A Framework for Instructional Planning and Discourse Modelling in Intelligent Tutoring Systems

This paper focuses on automatic tutoring using mixed-initiative dialogues in natural language. Two main issues are addressed: instructional design, i.e. how to plan and carry out a teaching session and, tutorial dialogue where two participants communicate in natural language to perform the educational task. We consider the tutor and the student as agents cooperating in a learning process but having their own goals. Under this assumption, the tutor initially states an instructional goal and plans a strategy to attain it. This strategy is refined in a dynamic way, so that the student’s interactions are always considered. A natural language interface has been implemented. The system understands student’s utterances, in Spanish, and builds the dialogue structure dynamically interchanging messages both with the student and the planner.