An object oriented model for expert system shell design

Abstract

A tuple-space-based object-oriented model for knowledge representation and interpretation is presented. This model is currently being applied to the expert system arena of knowledge-base-systems. The architecture addresses two important issues of knowledge base system design: the tradeoff between expressivity and tractability in knowledge representation, and knowledge-base-system and expert-system-shell domain independence. An architecture that falls somewhere between OPS5 and FOL in expressivity and tractability is defined. It trades some expressivity for tractability, and vice versa. Most expert system shells also tend to be somewhat domain restrictive (such as diagnostic and classification domains). Both the knowledge representation language and the inference strategy used contribute to this restrictiveness. Analysis of requirements for greater expressivity of both knowledge and inference strategy for representing domain independence, as well as for keeping intractability manageable, has resulted in an object-oriented language for defining knowledge-base management systems, as well as their instantiation. This language and its implementation are discussed.< >