Automatic Extraction of Developer Expertise

Cédric Teyton (1) , Marc Palyart (1) , Jean-Rémy Falleri (1) , Floréal Morandat (1) , Xavier Blanc (1)

1 LaBRI - Laboratoire Bordelais de Recherche en Informatique

\textbf{Context:~} Expert identification is becoming critical to ease the communication between developers in case of global software development or to better know members of large software communities. To quickly identify who are the experts that will best perform a given development task, both the assignment of skills to developers and the computation of their corresponding expertise level have to be automated. Since the real level of expertise is tedious to assess, our challenge is to identify developers having a significant level of experience with respect to a skill. \textbf{Method:~} In this paper we propose \xtic, an approach that takes up this challenge with the intent to be accurate and efficient. \xtic provides a language to specify skills. It also provides an automatic process that extracts skills and experience levels from source code repositories. Our approach is based on the idea that an expert has a high level of experience with respect to a skill. \textbf{Results:~} We have validated \xtic both on open source and industrial projects to measure its accuracy and its efficiency. The results we obtained show that its accuracy is between moderate and strong and that it scales well with medium and large size software projects. \textbf{Conclusion:~} \xtic supports the specification of a diversity of developer skills and the extraction of the expertise of these developers under the form of level of experience.

Domaines: Génie logiciel [cs.SE]

Cédric Teyton, Marc Palyart, Jean-Rémy Falleri, Floréal Morandat, Xavier Blanc. Automatic Extraction of Developer Expertise. 2014. ⟨hal-00965074⟩