The role of partial and best matches in knowledge systems

Partial matching is a comparison of two or more descriptions that identifies their similarities. Determining which of several descriptions is most similar to one description of interest is called the best match problem. Partial and best matches underlie several knowledge system functions, including: analogical reasoning, inductive inference, predicate discovery, pattern-directed inference, semantic interpretation, and speech and image understanding. Because partial matching is both combinatorial and ill-structured, admissible algorithms are elusive. Economical solutions require very effective use of constraints that, apparently, can be provided only by globally organized knowledge bases. Examples of such organizations are provided, and promising avenues of research are proposed.