1988 年 3 巻 3 号 p. 319-328
Whether an expert system remains usable or not depends on the maintenability of the knowledge stored in it. In order to allow non-programming users to refine the knowledge base easily, we assume the following: (l) Once a knowledge base has been installed, there are components that should not be altered under any circumstances, and components that should be refined continually in the knowledge base. (2) The latter components have structures and attributes that can be defined explicitly. Under these assumptions, we present meta-knowledge representation (FORM) and its interpreter (TAILOR). Form consists of the pattern, facet, and method. The pattern represents the structures of the knowledge base, the facet represents the attributes of the knowledge base, and the method represents how the contents of the knowledge base are altered. TAILOR transforms knowledge representation for experts into knowledge representation for inference engines.TAILOR allows an expert to describe knowledge because it serves as an application specific editor. Application of FORM and TAILOR to an IR expert system is also discussed.