This repository contains the code and data for paper On the Relevance of Cross-project Learning with Nearest Neighbours for Commit Message Generation (doi:10.1145/3387940.3391488)
nngen.py file at the root folder contains implementation of Simple-NNGen & EXC-NNGen.
The content of the files in "files/data" is described in the following table.
| File Name | Description |
|---|---|
| bleu_scores | scores of each method for tests from each dataset |
| [x].test.msg | messages generated by method [x] for tests from cleaned dataset |
| filtered-[x].msg | messages generated by the algorithm or copied from source for filtered data |
| msg-word-count | data for Figure 1 of paper. First line for training and second line for testing subset |
| nngen-avg-bleu-scores | bleu scores for messages generated by nngen for each origin |
| number-of-messages-from-the-same-repo | number of messages each algorithm selects from the same repo |
| [other/same/unknown]_repo.msg | nngen.test.msg & test.msg data divided due to the origin |
| [test/train].[diff/msg] | copied from the cleaned dataset |
| scores_per_origins | scores for messages selected by NNGen from each origin |
| [x].projectIds | repo_id for each message in [x] |
| repo-id-to-github.txt | maps repo_id to github repository |