Automated Retrieval of Artifacts Created during the Software Development Life-cycle

The number of failures of software projects not meeting the originally intended requirements are many. While often due to users and developers not sharing the same vocabulary, it is more often due to changes which are not reported or recorded somewhere along the development cycle. Software traceability (ST), is the process of tracking changes in the document corpus which are created throughout the software development life-cycle. There are known techniques, such as using traceability matrices, which attempt to solve the problem. Such mechanical methods are not only manually intensive, but they totally ignore the effects of synonymy and polysemy. Latent semantic analysis (LSA) is intended to avoid these latter effects and is largely used in the world of Information Retrieval (IR). In this report we apply LSA for the purpose of maintaining artifacts generated during the software development life-cycle and place greater emphasis than hitherto found in the literature, on term extraction in software code, something we call attribute weighting. We moreover present a software tool for the automation of the traceability process, including query refinement and show that the technique allows one to trace through the artifact corpus with the confidence that the set of artifacts affected by a change will be discovered