Verification · Code review · Reviewers learned from past reviews
Topic clustering
Cluster comments into themes with a minimum number of supporting discussions.
How to build it
- Embed each review comment as a vector with a sentence-embedding library.
- Cluster the vectors and drop clusters with fewer than the minimum number of independent threads.
- Record the cluster size and the threads behind each cluster.
Signals in a repository
- A clustering configuration with a min_cluster_size value in the analysis script.
- Cluster output files that list the threads supporting each theme.
In practice
No scanned repository runs a tool for this check. The steps above are the only guidance.