Proximity-based traceability: An empirical validation using ranked retrieval and set-based measures

The automatic generation of traceability links attempts to reduce the burden of building requirements traceability matrices (RTMs) that will be vetted by a human analyst before use in verification and validation tasks such as criticality assessment or change impact analysis. Information Retrieval (IR) techniques, notably the Vector Space Model (VSM), have been used with some success to build textual artifact traceability matrices. A limitation of the VSM is that it disregards word or term location and the relationship between words in the textual artifacts being traced. This paper presents a VSM enhancement with consideration for term location, validating it on four datasets using ranked retrieval and set-based measures. These two types of measures provide a more detailed comparison between the two traceability techniques.