Automatically detecting feature requests from development emails by leveraging semantic sequence mining

Mailing list is widely used as an important channel for communications between developers and stakeholders. It consists of emails that are posted for various purposes, such as reporting problems, seeking help in usage, managing projects, and discussing new features. Due to the intensive amount of new incoming emails every day, some valuable emails that intend to describe new features may get overlooked by developers. However, identifying these feature requests from development emails is a labor-intensive and challenging task. In this paper, we propose an automated solution to discover feature requests from development emails by leveraging semantic sequence patterns. First, we tag sentences in emails by using 81 fuzzy rules proposed in our previous study. Then we represent the semantic sequence with the contextual information of an email in a 2-g model. After applying sequence pattern mining techniques, we generate 10 semantic sequence patterns from 317 tagged emails that are randomly sampled from the Ubuntu community. We also conduct an empirical evaluation of their capability to discover feature requests from massive emails in Ubuntu and other four open source communities. The results show that our approach can effectively identify feature requests from these emails. Compared to existing baselines, our approach can achieve a better performance in terms of precision, recall, F1-score, AUC, and positive, with the average precision and recall for discovering feature requests from emails being 76% and 86%, respectively.