When to Release in Open Source Project?

As a rapidly growing number of open source projects adopt distributed version control systems, it becomes more crucial and challenging for release managers to oversee the software development, testing, deployment, and support during the software development life cycle. Manual monitoring and control of the lengthy commit history has been extremely tedious and effort consuming. This paper proposes an automated approach based on chronological commit logs to recommend release candidates for release managers by casting release tagging into a structured learning problem. It predicts whether a commit can be identified as a potential release point leveraging Conditional Random Fields (CRFs) model fueled by automatic feature extraction. Experiments on two open source datasets demonstrate promising results.