An approach to the dynamic verification of knowledge-based systems

Abstract

This paper describes a method of formal dynamic verification of knowledge-based systems. Almost all verification methods that have been proposed can validate the structural and logical integrity of knowledge-based systems, but they are not semantic. In other words, they are all static and syntactic. In addition, this paper adopts a dynamic approach in which verification is carried out in the reasoning process. This methid enables the detection of incorrect rules in the knowledge base using a simple debugging technique, which traces the execution of inference. Following the debugging algorithm, the user is asked several questions. To reduce the number of questions, knowledge is added to the system. In this way, some of the checks for correctness can be done by the system. Specifications are the judgement criteria of correctness, which are made up of some constraints. Application of the method to a power system operation is given. Benefits of this approach include: the ability to evaluate system's conclusions in situations where the experts may judge reasoning process, and the ability to predict the performance of the knowledge-based system in untested cases or to identify other cases that should be tested.