Monitoring sentiment in open source mailing lists: exploratory study on the apache ecosystem

Large software projects, both open and closed source, are constructed and maintained collaboratively by teams of developers and testers, who are typically geographically dispersed. This dispersion creates a distance between team members, hiding feelings of distress or (un)happiness from their manager, which prevents him or her from using remediation techniques for those feelings. This paper evaluates the usage of automatic sentiment analysis to identify distress or happiness in a development team. Since mailing lists are one of the most popular media for discussion in distributed software projects, we extracted sentiment values of the user and developer mailing lists of two of the most successful and mature projects of the Apache software foundation. The results show that (1) user and developer mailing lists carry both positive and negative sentiment and have a slightly different focus, while (2) work is needed to customize automatic sentiment analysis techniques to the domain of software engineering, since they lack precision when facing technical terms Keywords—Empirical Software Engineering, Sentiment Analysis, Mining Software Repositories, Mailing List Data