Using conceptual clustering for classifying reusable Ada code

One essential problem in reusability is classifying the reusable components. The library must be structured to facilitate the retrieving of code. Several approaches have been applied to this problem, but none addresses the evolution of software libraries. In this paper, we show how conceptual clustering can be used for dynamically classifying software components as the reusable library expands over time. Unimem is a conceptual clustering system that “learns” concepts by noticing similarities. We have adapted Unimem for developing and maintaining Ada libraries, where a large portion of the information necessary for our classifying method is directly extractable from the code.