State of the Art: Knowledge Acquisition and Modelling

Abstract

Knowledge acquisition and modelling play a leading role in the development of zknowledge-based systems (KBS). Traditionally, acquisition of knowledge was seen as the activity of transferring expertise from human experts or from technical documents in an abstract structure, in the form of models in the artificial world (Dieng, 1990). First generation knowledge-based systems were based on a knowledge transfer principle, i.e., knowledge was extracted from experts and transformed by knowledge engineers into a form that could be used by computers. Some representative systems are the STRIPS system (Fikes and Nilsson, 1993) and GPS (Newell and Simon, 1972). The expert’s natural language descriptions were transformed by the knowledge engineer into heuristic or empirical rules which mapped observable features of the problem to conclusions. This first generation of knowledge-based systems had a simple control structure and a uniform representation of knowledge in the form of associative production rules. A good overview and a comparison of acquisition methodologies used by first generation expert systems is found in (Neale, 1988). This overview discusses shortly the methodology KADS-I.