Knowledge Refinement using Rolls of Knowledge

Abstract

This paper describes a knowledge refinement method using functional information of knowledge. To build knowledge bases for expert systems, we must acquire knowledge from human experts. Knowledge acquisition support systems help us to extract and refine knowledge. If knowledge is represented in general knowledge representations (e. g. production rule), we can refine only its logical inconsistency. We must know and use each role of knowledge to detect its inconsistency and insufficiency. Therefore, we have developed a knowledge refinement method for knowledge represented in operation representation, called Expert Model. This representation has been derived from analyzed real knowledge bases in production rule form. It has seven functional types, which are selection, classification, ordering, combination, translation, input and output. To use these functional types, our method can refine inconsistency, over generalized knowledge, over specialized knowledge and insufficiency of knowledge efficiently.