Intelligent Collaboration Tools Lab

Reviewer Recommendation Revisited

The selection of appropriate reviewers for a code change can have a big impact on software quality. Automatic reviewer recommendation (ARR) aims to help developers find the most suitable reviewers. However, there are several gaps in current ARR research, which we aim to identify and help address.

Code review is an essential part of modern-day software engineering. The selection of appropriate reviewers is crucial for an efficient and effective review process, which ultimately impacts the overall quality of the software. Automatic reviewer recommendation (ARR) involves selecting the most suitable reviewers for a given code change. In recent years, the field of ARR has made significant advancements, but several gaps still exist. This project aims to facilitate the development of new ARR approaches by identifying and helping to address those gaps.

We performed a large empirical study of ARR literature to identify the most prominent issues and obstacles in ARR research and learn how they are addressed or mitigated by the community. Additionally, we built two open-source tools that solve some of the issues. The first of these is the MR-loader tool, which mines the most popular datasets of Gerrit-based projects without losing any data. The second is the Bat CoRe package, designed for building and testing new models. The package comes with implementations of popular baseline methods, optimizations, and tools for testing new models.

Currently, we are finalizing the results and finalizing our paper.

Participants

Vladimir Kovalenko
Farid Bagirov
Nikolai Sviridov
Pouria Derakhshanfar