DEFINITION OF A COMMUNICATION LAYER FOR EXPERT SYSTEMSzz

Sharing information and knowledge in environments where heterogeneous knowledge{based systems have to be integrated requires methodologies to systematically overcome the incompatibilities on various levels of communication. Besides overcoming incompatibilities on the syntax level (diierent representation languages) and diiering ontologies (diierent semantics) a unifying approach on the communication level is required. We present an approach where this level can be deened independently of the concrete realization of the system at hand. We describe the concepts of a tool that enables users to deene the communication skills of KBSs in a TELL and ASK fashion. KEYWORDS heterogeneous knowledge-based systems, knowledge sharing and reuse, communication layer, communication operators INTRODUCTION Due to its economical impact the reuse and sharing of existing solutions for former problems has been accepted and is even actively enforced in a large variety of engineering disciplines. Engineers are used to take advantage of existing conceptual solutions or, in case of constructing complex systems, use already realized subsystems in the manner of black boxes with well{deened functionalities. Also within computer science methodologies are currently being developed that provide techniques for reuseability, e.g. case{based reasoning in Arti-cial Intelligence (AI), reusable software modules in software engineering, etc. However, if we regard the construction of knowledge{based systems (KBS) as an engineering process and its result as an engineering artifact the reuse and sharing of such an outcome has only very recently been regarded as a problem of great scientiic interest and high economical impact and the available methods for this purpose are still in its infancy. There are probably many possible views on KBSs with respect to reuseability and sharing. One, which we will adopt here, is the standpoint of regarding a KBS as a black box that provides a well{deened problem{solving capability to the outside world. Hence, a large|in terms of complex functionality|KBS can be realized by using a set of already existing KBSs. Their output is integrated by the newly realized system. Consequently, the question arises how the integration can be achieved. In Distributed AI this question can be answered straightforward since all agents can be regarded as being identical along several dimensions, e.g. problem{solving capability, domain concepts, language at their interface, etc. Contrarily, we are faced with the situation where we want to integrate several heterogeneous systems, each providing a diierent problem solving capability to the overall system. Moreover, we regard a system with central management, i.e. we assume the existence …