Guilherme Bittencourt, Jacques Calmet, Karsten Homann and Anita Lulay Abstract
The intelligent behavior of a system is based upon its represented knowledge and inference capabilities. In this paper we report on a knowledge representation and reasoning system, developed at the University of Karlsruhe, called Mantra.
The system provides four different knowledge representation methods -- first-order logic, terminological language, semantic networks, and production rules -- distributed into a three levels architecture. The first three methods form the lowest level of the architecture, the epistemological level. The supported hybrid inferences as well as the management of knowledge bases form the second level, called logical level. Finally, the third level, the heuristic level, provides representation of procedural knowledge of a domain, and the introduction of ad hoc rules. This knowledge is represented in terms of production rules which are processed by a Ops5-like rule interpreter.
This paper mainly describes the introduction of this level into the hybrid system. The semantics of the knowledge representation methods of the epistemological level is defined according to a four-valued logic approach. This definition insures that all inference algorithms are sound, complete and decidable. The system has been implemented in Common Lisp using the object-oriented extension CLOS, and the graphical user interface was implemented in C with XToolkit.