Artificial intelligence and space

Abstract

A variety of knowledge acquisition (KA) techniques have proven useful for developing knowledge-based systems (KBSs). Such techniques typically have partially overlapping functionality with respect to the type of knowledge they may be used to acquire. Overlap between KA techniques means that parts of a knowledge base (KB) acquired using one technique may be refined and extended using some other technique; there is a need for knowledge interchange. This need is aggravated by the dynamic KBS development life-cycle, with frequent switches between life-cycle phases and techniques within each phase. Within this environment, it is very hard to predict exactly how knowledge interchange between acquisition techniques should take place. To address this problem, we have developed a knowledge reformulation technique and tool to support the knowledge engineer in performing knowledge interchange. The approach has two main types of functionality. First, it contains functions used for translation from and into knowledge acquisition KBs, using the Knowledge Interchange Format (KIF) as an intermediate language. The translation functions are based on grammars of KA techniques as well as approaches to formalize the knowledge acquisition KBs. We have developed translators for the well-known KA techniques card sort and repertory grid. The translators are written using the definite-clause grammar (DCG) formalism, which is based on Prolog. The second type of functions in the knowledge reformulation approach are denoted adaption functions. The adaption functions are used for KB organizing, structuring and editing. These functions are utilized by the knowledge engineer to modify KBs before and after translation. In this paper we present the knowledge reformulation technique and tool, along with the methodological basis for the approach. An example of how the Knowledge Reformulation tool (KRF) can be used is also included.