It is a well-known fact that the reuse of software greatly improves the productivity and reliability of software development. The most important issue concerned with the software reuse system is finding the appropriate components, which meet some given requirements. The early software reuse systems cost too much time and money for building software library and it became an obstacle for software reuse. To solve this problem, an approach for building software library automatically using information retrieval was suggested and eventually proved good performance. The text source used in the preceding research was mainly source codes or comments. However, these kinds of sources included useless information such as descriptions for the codes as well as information on the functions of software parts, and there was no way to automatically distinguish between them.
In this thesis, I propose a new approach for automatically extracting index from the text source. It is a specification based software reuse system. This approach is based on a hypothesis that the index extracted from specification would outperform the index extracted from comments, because specification only includes functional descriptions. Through experiments, the new system has proved that the index from specification performs 6% better than the index from comments.
