Experiences of Using Verification Tools for Maintenance of Rule-based Systems

Contents

This paper reports on the use of tools that statically analyze rule-bases in order to detect logical errors, such as redundancy, inconsistency, and incompleteness. The analysis uses a variant of the KB-Reducer algorithm (Ginsberg 1991). In contrast to accounts of similar tools, which seldom describe tools as used in the field, the experience described in this paper is derived from using tools to analyze a large suite of knowledge bases in production use at Boeing. The tools serve to augmenthe ongoing process of knowledge base maintenance. Common causes of errors detected are reported. This paper shows that making such tools practical is non-trivial but rewarding.