Classifying software components using design characteristics

Classifying software modules in a component library is a major problem in software reuse. Indexing criteria must adequately reflect the semantics of the components. This must be done without undue effort in either classifying the software, or developing "queries" to find candidates for reuse. We present an architecture for automatically classifying and querying software based on design information. We present a method for determining if indexing criteria are effective, and show results using a set of criteria automatically extracted from an existing collection of programs.