DEVA: sensing emotions in the valence arousal space in software engineering text

Existing tools for automated sentiment analysis in software engineering text suffer from either or both of two limitations. First, they are developed for non-technical domain and perform poorly when operated on software engineering text. Second, those tools attempt to detect valence only, and cannot capture arousal or individual emotional states such as excitement, stress, depression, and relaxation. In this paper, we present the first sentiment analysis tool, DEVA, which is especially designed for software engineering text and also capable of capturing the aforementioned emotional states through the detection of both arousal and valence. We also create a ground-truth dataset containing 1,795 JIRA issue comments. From a quantitative evaluation using this dataset, DEVA is found to have more than 82% precision and more than 78% recall.