F-MP - A FUZZY MATCH FRAMEWORK FOR RULE-BASED PROGRAMMING

Rule-based expert system shells have one important drawback in handlin g uncertain knowledge. It is the drawback that the style of their fuzz y reasoning process and their semantics both are not compatible with t hose of relational databases. On the other hand, production rule-based languages whose structure is similar to that of the databases fail to possess the fuzzy reasoning ability. Proposed in this paper is a fram ework to support a semantic based inexact match with Fuzzy Match Predi cate (F_MP). In a uniform way it allows matches including fuzzy lingui stic variables as well as fuzzy numbers. Our framework also adopts a d esign alternative to conform not only the semantics of its knowledge r epresentation but also its reasoning style to those of the relational framework. It is a natural consequence that such a design alternative entails a seamless integration of our work into the relational databas es. A major advantage of our framework is that it can be implemented o n top of the production rule-based languages without modifying their d iscrimination networks. That is mainly due to the minimal semantic gap between the framework and the languages. In this paper, we demonstrat e that: (1) F_MP is a uniform framework to provide the rule-based lang uages with fuzzy match facilities semantically enhanced, and that (2) its semantic conform well to that of the relational one. We also devel op a rule-evaluation mechanism well suited to the aims.