Using Developers Contributions on Software Vocabularies to Identify Experts

Developers choose identifiers to name entities during software coding. While these names are lexically restricted by the language, they reflect the understanding of the developer on the requirements that the entity is devoted for. In this paper, we analyze the use of such vocabularies to identify experts on code entities. For a real software development, e-Pol (Management Information System for Federal Police of Brazil), we observed around 30% of its code entities has more than 0.3 of similarity with at least one developer vocabulary. We propose an approach to catch this potential expertise that vocabularies carries on. Also, we built an oracle of source code entities per developer that allowed us to assess our approach accuracy compared with two others ones: based on commit and based on percentage of modified Lines of Codes. One advantage of our approach is to disregard changes in formatting or indentation of source code as acts of expertise acquisition. We achieve an accuracy ranging from 0.16 to 0.32 depending on the assumed period of developers´ contributions and the top-k experts we are interested on. These results confirm similarity between vocabularies might be explored to point out code experts. Moreover, for orphaned entities, expertise approach based on vocabularies can recommend among current team members one whose vocabulary is closest to the entity.