Showing papers on "Specialization (logic) published in 1992"

The “descriptive” component of a hybrid knowledge representation language

Abstract: We describe one component of a "hybrid" Knowledge Representation Language (KRL) used for tile development of Large Knowledge Bases (LKBs). Tiffs hybrid language involves two dif- ferent aspects, the "descriptive" and the "definitional." The representation of the elementary events occurring in the real world (descriptive data = "Snoopy is CharUe Brown's beagle") is organized around "semantic predicates" (with "roles" and "arguments"); this gives rise to units called "pred- icative occurrences." The single predicative occurrences can be combined using logical, causal etc., relationships, giving rise to complex conceptual constructions (*'binding occurrences"). "Abstract" conceptual units ("templates") which describe the expected properties of the "concrete" predicative and binding occurrences are placed in a specialization hierarchy (H.TEMP) characterized by the inlmritance of properties nmd behaviours; the concrete "occurrences" constitute the "leaves" of this hierarchy. On the other hand, the "classes ~ representing the "general categories" of all the basic entities of the application domain which appear in the predicative occurrences are defined in terms of their own specialization hierarchy, tl_CLASS (this is deJiaitional data = "A beagle is a sort of hound / a hound is a dog ... "); the concrete "instances" of the defined classes are the '*leaves" of this second hierarchy. The "delitfitional component" is not discussed in this article. The main difference between this KRL and other recent "hyl~rid" hmguages is that the "descriptive component" is not a first order predicate calcuh,s Imlguage, but a highly structured setnmttic network.

Representation of Fragmentary Multilayered Knowledge

Abstract: Formalization presupposes ‘precisification’. A formal representation, therefore, cannot account for all relevant aspects of imprecise domain knowledge. In this paper we present a methodology for dealing with this problem. In an imprecise domain, part of the expertise is to know the realm within which knowledge may be faithfully specialized. In a computer reasoning system, such expert knowledge can be reproduced as a metatheory for proposing and reasoning with formal object theories, each representing one particular specialization. Metaknowledge of this kind will however most often also be imprecise and the expertise on how it may be specialized resides then at the metametalevel, etc. We show that logic provability and upward reflection are adequate means for representing such hierarchical domain knowledge and the dependencies in it between adjacent levels.