The management of acquired knowledge in expert systems

This paper is concerned with the methods and strategies behind the acquisition of new knowledge in an expert system and its subsequent management. The acquired knowledge is assumed to be stored in a database which is not part of the knowledge base of the system. By the process of analogical inference the acquired knowledge is used to produce results, partial or whole, that depend on the measure of similarity between the problem set and the facts stored in the database. The paper argues for the application of inference mechanisms on the acquired knowledge so that the outcome of such inference can be used as a heuristic for reducing the search space relating to the given problem. There are three different topics discussed here: analogical reasoning; inductive inference; and a combinational learning strategy. A combination of these can be used to minimize the number of production rules inferred.<>