Latent Co-development Analysis Based Semantic Search for Large Code Repositories

Distributed and collaborative software development has increased the popularity of source code repositories like GitHub. With the number of projects in such code repositories exceeding millions, it is important to identify the domains to which the projects belong. A domain is a concept or a hierarchy of concepts used to categorize a project. We have proposed a model to cluster projects in a code repository by mining the latent co-development network. These identified clusters are mapped to domains with the help of a taxonomy which we constructed using the metadata from an online Question and Answer (Q&A) website. To demonstrate the validity of the model, we built a prototype for semantic search on source code repositories. In this paper, we outline the proposed model and present the early results.