Mining software repositories towards understanding code change properties to guide program repair
Patching is a common activity in software development. It is generally performed on a source code base to address bugs or add new functionalities. In this context, given the recurrence of bugs across projects, program repair tasks can be improved by mining and learning templates of candidate changes from program code and version control histories. In this project, the aim will be to build an automated program repair-adapted representation of code that will be leveraged to associate the change intention to various artifacts such as bug reports, test cases. Concretely, we will explore various code embedding approaches towards learning a deeper semantic representation of code in order to more accurately and efficiently reason about the recurrence of code changes.