Traceability Recovery Using Numerical Analysis

The paper proposes a novel information retrieval technique based on numerical analysis for recovering traceability links between code and software documentation. The results of a reported case study demonstrate that the proposed approach significantly outperforms two vector-based IR models, i.e., the vector space model and latent semantic indexing, and it is comparable and sometimes better than a probabilistic model, i.e., the Jensen-Shannon method. The paper also discusses the influence of each method with the specific artifact type considered and the artifact language.