A Deductive Pattern Matcher

Abstract

This paper describes the design of a pattern matcher for a knowledge representation system called LOOM. The pattern matcher has a very rich pattern-forming language, and is logic-based, with a deductive mechanism which includes a truth-maintenance component as an integral part of the pattern-matching logic. The technology behind the LOOM matcher uses an inference engine called a classifier to perform the matches. The LOOM matcher is more expressive and more complete than previous classification- based pattern-matchers, and is expected to be significantly more efficient.