Classify a commit's data-refactoring type from its change, not just the message: combine message keywords with diff signals (file-level ops + column-set delta from analysis/inspect_commit.py). This is the baseline classifier.
Score against the hand-labeled gold set (tests/gold_set.csv).
Blocked by labeling the gold set. Followed by evaluation.
Classify a commit's data-refactoring type from its change, not just the message: combine message keywords with diff signals (file-level ops + column-set delta from
analysis/inspect_commit.py). This is the baseline classifier.Score against the hand-labeled gold set (
tests/gold_set.csv).Blocked by labeling the gold set. Followed by evaluation.