1. Motivation A Framework for an Efficient DBS-Support of Knowledge-Based Systems

During the last few years a variety of Knowledge Based Systems (KS) has been developed as an application of techniques from the area of Artificial Intelligence (AI). When modeling real world applications, these systems are faced with problems of managing large amounts of knowledge, since virtual memory sizes are not large enough to store the corresponding Knowledge Bases (KB). Realizing that existing Database Systems (DBS) provide features to fulfil this kind of requirement, research efforts have been done with the purpose of coupling KS and DBMS. So, the knowledge engineering tool KEE [FK85], for example, has been extended with a new component ( the so-called KEEconnection [In87]), which enables it to store facts in an external DB to be managed by an independent relational DBS. The coupling approach may perhaps solve KS limitations concerning virtual memory size, but it fails, however, to support KB management for several other reasons [Ma88b]. For example, when KB are maintained on secondary storage devices, operations on knowledge (e.g. inference) are so computationally intolerable, that coupling DBMS and KS yields very low performance [HMP87]. The solution to the DB deficiencies to support KS is to develop a new generation of systems, the so-called Knowledge Base Management Systems (KBMS), aimed at the modeling and manipulation of knowledge as well as at its maintenance in very large, and possibly distributed, KB. Following this shift of viewpoint, we implemented at the University of Kaiserslautern a multi-layered prototypical KBMS, called KRISYS, which supports the above mentioned demands in an effective and efficient manner [Ma88b]. In this paper, we concentrate on the KRISYS component responsible for the KB management, thereby describing the framework provided by the system for an efficient processing of knowledge. 2. The Framewor