Compression

Abstract

The utility of a search control rule in PRODIGY is highly dependent on its match cost, the time cost of matching its left-hand side. Whenever PRODIGY checks to see whether a control rule is applicable, this time cost is incurred. In order to reduce the match cost of the rules that it learns, PRODIGY uses a process called compression. Given a description, the compression module (the COMPRESSOR) attempts to find a semantically equivalent description that is less expensive to match. The process is similar to the rather vague notion of “simplification”. However compression, unlike simplification, is specifically intended to reduce a description’s match cost. In practice, compression also tends to make descriptions more concise and thus easier to read, but this is considered a fortuitous side effect.