Advances in local student modeling using informal fuzzy reasoning

This paper presents an approach to local student modeling in mathematics intelligent tutors. A knowledge representation is developed that stores separately the semantic and structural information needed to represent math word problems. Based on this representation, and the assignment of weights to semantic labels for problem sets, the student is allowed considerable flexibility in the development of solutions. The technique used in matching student solutions to acceptable solutions is based on imprecision or fuzziness, i.e. exact matches to stored solutions are not required. Moreover, this imprecise approach, together with a ''collapsing'' of the tree of all possible solutions, substantially reduces storage and search requirements, addressing some combinatorial explosion and speed problems of modeling. This method can support a variety of instructional strategies. (C) Academic Press Limited