Software storage structuring and retrieval remain a major challenge to the widespread adoption of software reuse. An approach that can facilitate the automatic structuring of software components libraries is presented. Based on automatic indexing and self-organising map (SOM) technologies, key features associated with software components can be identified and organised in a very simple way that makes their distance relations geographically explicit on the two-dimensional output layer of the SOM. The proposed approach was applied to a collection of UNIX commands to evaluate its retrieval effectiveness. Preliminary results were encouraging and showed improvement on both recall and precision, with substantial reduction in the amount of human effort required in the process.
IEEE Proceedings Software Vol. 148, Issue 2, p. 45-55