Design of a General Planmatcher for Diagnosing Student Programs

Several Intelligent Tutoring Systems (ITSs) for teaching imperative languages have been built up until now. However, most of them only teach one particular programming language. Moreover, they lack modular composability, that is, existing components of the available ITSs can hardly be reused to create a new ITS. Furthermore, introducing new problems or feedback requires a lot of effort and expertise. We propose to define the feedback and problems at the application level in order to provide a more reusable and extensible ITS. We have built an ITS shell which can generate ITSs for teaching different imperative programming languages. The ITSs generated by our shell is composed of an instruction model, domain model and a student model. In order to use the shell system for instructional experimentation, the domain model and the instruction model are highly orthogonal. This paper handles about the Planmatcher which is a basic component of our ITS shell. The Planmatcher is used for matching a given problem specification with a student program written in an arbitrary imperative programming language. A new general problem specification language (PSL) is introduced to gain the necessary orthogonality. Matching is done using object-oriented algorithms provided in a reusable and extensible framework for generic programming plans. Because of the high level of orthogonality of the components new problems and feedback can be easily added or modified on the application level. keywords: intelligent tutoring system shell, expert systems, knowledge representation, planmatching, feedback generation, object-oriented software development