Control of inconsistency and redundancy in PROLOG-type knowledge bases

摘要

Because of the incremental and piecemeal nature of its construction, logical inconsistency and redundancy can be built inadvertently into a knowledge base. This paper discusses a methodology for analyzing the contents of a PROLOG-type knowledge base and for eliminating inconsistent and redundant logical elements. It introduces a graphical representation, the goal-fact network, of the logic required to infer a goal and describes the identification of inconsistency in that network. Three increasingly general alternatives, Boolean algebra, the Karnaugh map, and the Quine-McCluskey algorithm, are presented as tools to identify and to eliminate redundancy.