Uma abordagem para classificação de comentários e análise de impacto no tempo de resolução em relatórios de bugs: um estudo de caso no Bugzilla.

Abstract

Bug reports are important sources of information about problems that occur during software development. Several lines of research are focused on understanding these reports and how they can help the software development community. In this study, we propose an investigation of bug reports with an emphasis on the analysis of their comments. We seek to obtain information about the influence of comments on their resolution time. To this end, we define a taxonomy of comment classes built from a qualitative analysis of textual content and associated attachment types, with the objective of representing the most recurrent forms of contribution observed during the bug report lifecycle. Based on the taxonomy, Machine Learning models were trained and evaluated for the automatic classification of comments. The experiments demonstrate that the models are able to classify comments consistently among the proposed classes, allowing the identification of patterns between comment types and bug resolution times. The results indicate that certain classes of comments are associated with faster resolutions, while other classes are related to longer resolution times, highlighting the relevance of comment classification as a supporting tool for the analysis of the software maintenance process.