Learning to Update Natural Language Comments Based on Code Changes

Code and datasets for our ACL-2020 paper "Learning to Update Natural Language Comments Based on Code Changes" which can be found here.

If you find this work useful, please consider citing our paper:

@inproceedings{PanthaplackelETAL20CommentUpdate,
  author = {Panthaplackel, Sheena and Nie, Pengyu and Gligoric, Milos and Li, Junyi Jessy and Mooney, Raymond J.},
  title = {Learning to Update Natural Language Comments Based on Code Changes},
  booktitle = {Association for Computational Linguistics},
  pages = {1853–1868},
  year = {2020},
}

Download generation and update data from here.

1. Create a directory named generation-models in the root directory

mkdir generation-models

2. Train the comment generation model:

python3 comment_generation.py -data_path public_comment_update_data/full_comment_generation/ -model_path generation-models/model.pkl.gz

3. Evaluate the comment generation model:

python3 comment_generation.py -data_path public_comment_update_data/full_comment_generation/ -model_path generation-models/model.pkl.gz --test_mode

4. Create a direction named embeddings in the root directory

mkdir embeddings

5. Save pre-trained embeddings from the comment generation model to disk:

python3 comment_generation.py -data_path public_comment_update_data/full_comment_generation/ -model_path generation-models/model.pkl.gz --test_mode --get_embeddings

6. Create a directory named update-models in the root directory

mkdir update-models

7. Download resources folder from here, and move it to the root directory.
8. Train the comment update model (i.e., edit model):

python3 comment_update.py -data_path public_comment_update_data/comment_update/ -model_path update-models/model.pkl.gz

9. Evaluate the comment update model (i.e., edit model):

python3 comment_update.py -data_path public_comment_update_data/comment_update/ -model_path update-models/model.pkl.gz --test_mode --rerank