Recovering Documentation-to-Source-Code Traceability Links using Latent Semantic Indexing

Author(s): Marcus, A., Maletic, J.
Venue: ICSE '03
Date: 2003

Type of Experiement: Case Study
Sample Size: 2



This paper repeats the experiments of Antoniol et al. with a different information retrieval technique. Specifically, this paper compares latent semantic indexing (LSI) to the methods studied by Antoniol et al. (probabilistic and vector space models) using the same data sets as Antoniol et al. (LEDA and Albergate). One difference between the studies is that the authors of this paper traced documentation to source code---the opposite of Antoniol et al. who traced source code to documentation.

The results of the two case studies are very similar. LSI exhibits similar recall and precision numbers for all cut points used in the study. (A cut point limits the number of documents returned by a query.) The techniques employed by the authors require less processing of the source code than the techniques of Antoniol et al.