An object-oriented knowledge base for multi-domain expert systems

Abstract

With the current technology of "deep " expert systems, domain knowledge can be modeled, using an object-oriented representation, separately from the strategic knowledge required to make inferences on the domain. Where such domains involve the construction of some hierarchy or network, a generic inference strategy can be devised to search the domain knowledge for plausible solutions. Thus, our main concern is in building an appropriate object-class structure that accurately models the problem domain. Rules are merely written to customize the expert system according to the uniqueness of each problem domain. Thus, a common framework for user interaction can be devised for domains of the same "type." By accessing the various domains stored in various knowledge bases, an individual is able to get the benefit of "expert advice" about several areas, using a somewhat similar interface, varying only according to context and terminology. Hence, the objective of this paper is to show how deep knowledge can be used to effectively model the descriptive information of a problem domain without the need to develop and maintain a large rule network. Consequently, we suggest a generic inference strategy that reasons on the deep knowledge via the application of heuristic functions. © 1995.