A graphic explanation environment for expert systems

The explanation facility is an essential part of an expert system. In general, there are three kinds of explanations: explanations during user input, explanations of knowledge base contents (static explanations). The last two types of explanation are considered. The authors show how to generate graphic explanations based on bipartite graphs. This explanation environment is able to answer both questions about the (static) knowledge base and the (dynamic) inferences. Implementation issues are considered. The inference engine writes a log (history) which reflects the state of dynamic knowledge during problem solving. Successful and unsuccessful rules are entered separately into the log in the sequence they are considered. In order to minimize the time for generating answers, the authors use a frame-based knowledge representation for the variable-rule interrelations which are extracted automatically from the rule base and the log. As examples, they describe the basic 'WHY' and 'WHY-NOT' questions and the algorithms used.<
>