GitHub Issue Mining to enhance Developer Profiling: A pipeline based on NLP tasks

The analysis of software artifacts plays a crucial role in supporting processes aimed at profiling developers. These artifacts, generated through interactions on versioning platforms such as GitHub, incorporate both structured and unstructured data. While existing literature primarily focuses on structured data from commit events, this study explores the potential of leveraging textual data to enhance profiling tasks, analyzing the developer interaction with other team members and/or users. In particular, we propose a Natural Language Processing (NLP) pipeline that utilizes topic modeling and sentiment analysis techniques to process GitHub Issue textual data. This approach has been evaluated using one of the most detailed just-in-time defect prediction datasets. Although the proposed pipeline has the capacity to support a general design of developer-related features that can be thoroughly analyzed using a wide range of approaches, in this paper, we employ clustering approaches for this purpose. The results demonstrate that features related to textual information provide a richer and more accurate representation of developer activities, providing valuable insights into data mining operations within software repositories.