A technique for knowledge base verification

A technique for detecting inconsistencies in, and incompleteness of, a knowledge base is proposed. The technique is based on modeling a knowledge base using a predicate/transition (Pr/T) net representation. Inconsistency and incompleteness patterns in a knowledge base are then defined with respect to the Pr/T net model and are identified by using a syntactic pattern recognition method. This technique can be included as part of a knowledge acquisition process in any rule-based system. It can detect major types of inconsistencies and incompleteness. The use of the technique can be easily automated. An example of its use is presented.<<ETX>>