Incorporating concept-based match into fuzzy production rules

Abstract

F_TP (fuzzy template predicate) is proposed as a template to incorporate concept-based match into fuzzy production languages. A thesaurus augmented in F_TP supports the concept-based match, which is more sophisticated than previous fuzzy match mechanisms. Membership functions for fuzzy linguistic variables and fuzzy numbers are used as an interface to the thesaurus. F_TP also has a self-refining facility to reformulate itself by exploiting user input or integrity constraints in the knowledge base. F_TP is used mainly in the left-hand side LHS of a production rule to participate in matches made between patterns and fuzzy facts. For the seamless integration of the rule into a relational database, the structure of the fuzzy facts LHS refers to is conformed to the relational framework. We also develop a rule evaluation mechanism supporting the F_TP construct. In the rule evaluation, a modified Rescher-Gaines translation rule is used for fuzzy implication rather than the well-known compositional rule of inference. The modification is also directed toward conforming the semantics of our rule evaluation to that of the relational framework.