Combining object-oriented representations of knowledge with proximity to conceptual prototypes

Abstract

A framework for knowledge representation that combines the fuzzy reasoning of systems and object-oriented databases is suggested. The use of objects to represent knowledge has become popular. However, this organization of knowledge, as a classification of entities by means of their attributes and their characteristic operations, returns to a traditional view of the formation of concepts (H. Gardner, 1985). This view, that conceptual categories can all be defined in the crisp way that mathematical concepts are defined, is not plausible for many real-world examples, and the idea of categories as formed from a clustering of data around a conceptual prototype, with an associated nearness measure, was substituted in its place (E. Rosch, 1978). A system that combines these two apparently distinct means of representation is described. Machine learning techniques are applied to the formation of suitable metrics for concepts.<>