An Improved VSM-based Post-Requirements Traceability Recovery Approach Using Context Analysis

Automatically generating traceability links between software development artifacts existing throughout systems development life cycle, is becoming ever more important for requirements traceability. It remains an open software engineering challenge, especially for legacy systems, when the demand for minimizing human intervention is considered. The Vector Space Model (VSM), a notably known information retrieval technique, attempts to remedy the situation by reducing the required manual effort. One limitation of VSM is its low-level performance in practice, which can be improved by involving human intervention in the requirements traceability process earlier. The contribution of this paper is to present an improved VSM-based post/requirements traceability recovery approach by using a novel context analysis. This is done by firstly removing redundant information in the search space of the artifacts wrt a requirement, and then using both requirement and context queries to refine the results given by the standard VSM. In this way, the subsequent artifacts from the source requirement are more likely to be retrieved in the recovery process. Our approach is evaluated by using two chosen datasets (i.e., eTour and iTrust), of which results show that the proposed approach can achieve better performance in terms of discovering more true trace links and obtaining higher quality lists of traceability links than the standard VSM.