DESIGN SCHEMES FOR RULE-BASED SYSTEMS

The design of rule-based systems is often plagued by errors and anomal ies. In this paper, we assume that a way to abstract the domain knowle dge is to use goals. The problem solving model of the system is in ter ms of goal inferences that occur in the rule base level using sets of rule sequences called paths. The mapping between goals and their repre sentation in a rule base is one-to-many and results in several design schemes for rule-based systems: the various design schemes impose diff erent mapping restrictions for representing a goal in a rule base. Thu s, the applicability of a design scheme and the relationships between the various design schemes are important factors that need to be consi dered before choosing a scheme for development. We explore one specifi c relationship, called inheritance, between the design schemes. The pr os and cons of using two schemes D-A and D-B, that have an inheritance relationship between them are analyzed for development versus mainten ance. The paper also provides metric measures to facilitate general co mparison between design schemes. Finally, we outline an application of inheritance for automatically transforming a rule base from one schem e into another.