Using a Description Classifier to Enhance Deductive Inference
Robert M. MacGregor
USC/Information Sciences Institute
4676 Admiralty Way
Marina del Rey, CA 90292
[email protected]
Abstract
The representation languages found in many expert system shells are
hybrids composed of a frame language and a rule language.
Unfortunately, the frame and rule components in these systems are not
well integrated, and as a result they miss important classes of
inferences. In place of frames, LOOM combines a description
language with a rule language, and uses an inference engine called a
classifier to achieve a successful integration of frame-like knowledge
and rule-like knowledge. LOOM's ability to reason with descriptions
enables us to implement a broader range of capabilities than those found
in the current generation of expert systems shells. For example, LOOM
is able to detect inconsistencies in a rule base, to match against
partially-specified (not fully-grounded) instances, and it implements a
generalization of the traditional (e.g., CLOS-like) conception of
object-oriented method dispatching.
In Proceedings Seventh IEEE Conference on AI Applications,
pp. 141-147, 1991.
The full paper is available in postscript. Get Postscript. (7pp)
Back to Paper List