Some Problems and Proposals for Knowledge Representation

A new theory of knowledge representation is proposed, called Cognitive Representation Theory (CRT), which eliminates the frame/slot distinction found in frame-based languages and incorporates as representational entities notions reminiscent of natural language metaphoric and metanymic relationships.

Abstract

A new theory of knowledge representation is proposed, called Cognitive Representation Theory (CRT). CRT encompasses representational ideas that have emerged from work in semantic networks, frames, frame semantics, and Conceptual Dependency. The theory attempts to meet certain desiderata for a meaning representation, namely, the principles of adequacy, interpretability, uniformity, economy, and, in particular, cognitive correspondence. Motivated by these principles, the theory eliminates the frame/slot distinction found in frame-based languages (alternatively, node/link distinction found in semantic network-based systems). In its place is a new notion called the absolute/aspectual distinction. In addition, the theory incorporates as representational entities notions reminiscent of natural language metaphoric and metanymic relationships. This is done through a mechanism called views. The theory allows for the representation of some ideas that in the past have only been represented procedurally, informally, or not at all. An implementation of much of CRT, called KODIAK, has been created, and used in a number of experimental AI systems