Knowledge-based discourse generation for an intelligent tutoring system
Yuemei Zhang · 1992
The goal of this research is to generate interactive discourse in a machine tutoring environment. Discourse in this environment must be coherent, integrative, and productive. In order to generate such tutoring discourse, this research must confront the following issues. The first is how to represent domain knowledge and contextual information to support tutoring discourse generation. The second issue concerns what strategies should be used to plan tutoring discourse based on the knowledge representation produced as a result of research on the first issue. The third issue is how to implement the planned discourse in natural language. This research attempts to study these problems and has achieved the following results. First, the issues in the process of generating tutoring discourse have been identified and investigated. Before this research, very little work had been done in this field. Second, detailed analysis of human tutoring sessions has been carried out. This led to the definition of the tutoring discourse as coherent, integrative, and productive discourse. It is also concluded that the student knowledge status, the dialogue history, and the domain knowledge determine what kind of discourse should be generated. Third, the domain knowledge base has been designed and implemented. It can be used not only to solve the domain problems and generate multiple problem solving paths, but also to provide support and explanation knowledge for discourse generation. The concept of tutoring context space was first proposed to facilitate discourse generation. It is believed that this context space can also be used to analyze discourse generated by students. Fourth, based on previous research on discourse schemata and rhetorical structures, interactive tutoring discourse schemata have been first proposed and developed in terms of discourse purposes, functions, and contents. These schemata can be used to generate tutoring discourse goals and plan discourse structures. Fifth, a sentence generator has been designed and implemented based on Lexical-Functional Grammar.