Detecting topic-based communities in social networks: A study in a real software development network

Abstract

In social network analysis, a key issue is the detection of meaningful communities. This problem consists of finding groups of people who are both connected and semantically aligned. In the software development context, identifying communities considering both collaborations between developers and their skills can help to address critical elements or issues in a project. However, a large amount of data and the lack of data structure make it difficult to analyze these networks’ content. In this paper, we propose a framework for detecting overlapping semantic communities and their influential members. We also propose an ontology to extract topics of interest through tag enrichment in a Q&A forum. An evaluation was conducted in a large network of software developers built with Stack Overflow’s data, showing that the proposed framework and ontology can find real communities of developers. The evaluation indicates that their members are semantic aligned and still active in the detected topics of interest, and the quantitative analysis showed that the detected communities have high internal connectivity.