EVALUATING THE USEFULNESS OF Q&A DEVELOPMENT PLATFORMS WITH TEXT MINING APPROACHES: THE EXEMPLARY CASE OF STACK OVERFLOW

Abstract

In recent years, Question and Answering (Q&A) platforms – multi-purpose as well as specified ones -became more prominent as a fast and mostly free accessible source of knowledge over various domains. One such platform focused on the area of software development within different areas is Stack Overflow, which was made publicly available in 2008. Since then, the platform became prominent as resource of knowledge for different kind of development tasks. To the best of knowledge, this work isthe first to suggest investigating Q&A development platforms by using Text Mining approaches, specifically supervised binary sentiment analysis to determine of the platform's usefulness as source of knowledge for the Text Mining development community. Facing different challenges like a severely skewed and small data sample and in addition a very context-specific style of language lacking diverse emotional expressions, the supervised sentiment analysis approach did not turn out to be sufficient for predicting the platforms usefulness with a performance minimal better than random choice. As an alternative indicator of platforms usefulness, a lexicon-based approach in combination with metadata variables was applied, which was originally used as initial labelling algorithm for the dataset. According to the results, Stack Overflow can be regarded as a platform with potential of being useful for the Text Mining development community, although further research is necessary.