ARSENAL-GSD: A framework for trust estimation in virtual teams based on sentiment analysis

Abstract

ContextTechnology advances has enabled the emergence of virtual teams. In these teams, people are in different places and possibly over different time zones, making use of computer mediated communication to interact. At the same time distribution brings benefits, it poses challenges as the difficulty to develop trust, which is essential for team efficiency. ObjectiveIn this paper, we present ARSENAL-GSD, an automatic framework for detecting trust among members of global software development teams based on sentiment analysis. MethodsTo design ARSENAL-GSD we made a literature review to identify trust evidences, especially those that could be captured or inferred from the automatic analysis of data generated by members’ interactions in a versioning system. We applied a survey to validate the framework and evidences found. ResultsOn a scale of 0–9, evidences were evaluated as having importance greater or equal to 5.23, and the extraction techniques used to estimate them were considered as good enough. Regarding differences between subjects profile, no difference was found in responses of participants with theoretical knowledge/none and those with medium/high knowledge in GSD, except for the evidence mimicry, which was considered more important for the group of participants with medium/high knowledge in GSD. ConclusionWe concluded that our framework is valid and trust information provided by it could be used to allocate members to a new team and/or, to monitor them during project development.