Knowledge refinement in diagnostic expert systems

This paper presents a new methodology for the incremental refinement of a knowledge base consisting of inductive rules. The novelty of the approach resides in the use of a body of deep knowledge for guiding the process of rule refinement, even in case this deep knowledge is too complex or not specific enough to deductively generate classification rules. Justifications derived from the deep knowledge can be exploited in order to localize failures and to propose changes oriented to improve the knowledge base. If only incomplete justifications are found, statistical evidence is used to select those which are likely to be the most reliable ones; justifications are also used to evaluate proposed changes in the knowledge base; this is done by reasoning on many examples and counterexamples at the same time.