Abstract
This paper introduces the principal concepts in the organization and operation of the logic based knowledge processing system, called CK-LOG (A Calculus for Knowledge in Logics). CK-LOG uses the frame based system MDS (the Meta Description System) for knowledge representation and for modeling world states. It uses an inference engine based on Natural Deduction for stating and solving problems. As a knowledge processing system CK-LOG has several capabilities, which are new to the technology of knowledge representation systems: CK-LOG has special facilities to represent and reason about actions and their time dependencies. Actions that occur in a world state may create or destroy objects in the world or modify their properties, or prevent or support other actions. The effects of actions are described in CK-LOG using modal operators like CREATE, DESTROY, PREVENT, SUPPORT, KEEP, etc. These operator expressions are also used to represent and reason about possible worlds that the actions might lead to. Most significantly, CK-LOG is a logic-based knowledge processing system, just as PROLOG is logic based programming system. CK-LOG uses a three valued logical system with truth values T (true),? (Unknown) and F (false) to build partial models of world states, and the two valued logic's system of T and F in its theorem proving System. The use of the three valued logical system in its models of world states enables CK-LOG to do problem solving in the context of incomplete information about world states. The theorem proving system of CK-LOG uses a variant of the calculus of sequents first proposed by Kanger (which itself is a variant of Gentzen's system). The two variations in CK-LOG are, (i). the use of a new algorithm called the mating algorithm for testing proof terminations, and (ii) the use of specialized inference rules for reasoning about modal expressions using the possible world semantics.. The mating algorithm gives the theorem proving system of CK-LOG several new capabilities: to identify information that is pertinent to a given problem and retrieve it from its knowledge base, to update its models of possible worlds during the problem solving process based on the findings of the theorem proving system, to use these models of world states to test proof terminations, and to generate hypotheses during the problem solving process that are based on unknown information. These various features of CK-LOG are described here. The paper concludes with a discussion of the logic of frames as used in CK-LOG and establishes a condition called locality condition as a sufficient condition for creating knowledge representations with requisite completeness.