Open-Source Software (OSS) community members report bugs, request features or clarifications by writing messages (in unstructured natural language) to mailing lists. Analysts examine them dealing with an effort demanding and error prone task, which requires reading huge threads of emails. Automated support for retrieving relevant information and particularly for recognizing discussants’ intentions (e.g., suggesting, complaining) can support analysts, and allow them to increase the performance of this task. Online discussions are almost synchronous written conversations that can be analyzed applying computational linguistic techniques that build on the speech act theory. Our approach builds on this observation. We propose to analyze OSS mailing-list discussions in terms of the linguistic and non-linguistic acts expressed by the participants, and provide a tool-supported speech-act analysis method. In this paper we describe this method and discuss how to empirically evaluate it. We discuss the results of the first execution of an empirical study that involved 20 subjects.
Official version of the paper.