- For Windows, run the following cmd to add python to sys.path and check
set PYTHONPATH=%cd%
echo %PYTHONPATH%- While for Linux, run cmd below
export PYTHONPATH=$(pwd)- Dependencies
jdk >= 1.8
nltk == 3.6.5
scikit-learn == 0.24.2
lightgbm == 3.3.2
nlg-eval == 2.3.0
tqdm == 4.62.3 - Output CCI samples' category prediction labels with the original data.
cd source
python HebCup.py ../Dataset/train_clean.jsonl 1 true
python HebCup.py ../Dataset/valid_clean.jsonl 1 true
python HebCup.py ../Dataset/test_clean.jsonl 1 true*This step will output the corresponding three dataset with labels (e.g., train_clean_1.jsonl) for subsequent steps.
- Discuss base classifiers on 10-fold cross-validation. Select one of the below choices to run. You can also add other classifiers by yourself to participate the comparison.
cd Ours
python eval_clfs.py lightGBM | naive_bayes | mlp | lstm | cnn- Train and evaluation for category classifiers
cd Ours
python classification.py lightGBM <suffix>*This process will output two files (HCUP_test<suffix>.jsonl and CUP_test<suffix>.jsonl) that are used for the following two steps.
- Evaluate the performance of HebCup Side
cd source
python HebCup.py ../Dataset/splitted_test_sets/lightGBM/HCUP_test<suffix>.jsonl 1 falsePlease refer to the original repo of HebCUP for detailed doc of HebCUP.
- Evaluate the performance of CUP side
To do this, you can send the CUP_test<suffix>.jsonl to the repo of CUP to evaluate the performance of CUP side with the metrics in this repo. Please refer to the original repo of CUP for detailed doc of CUP.
- If you find this paper useful and would like to cite it, the following would be appropriate:
@article{yang2022significance,
title={On the Significance of Category Prediction for Code-Comment Synchronization},
author={Yang, Zhen and Keung, Jacky Wai and Yu, Xiao and Xiao, Yan and Jin, Zhi and Zhang, Jingyu},
journal={ACM Transactions on Software Engineering and Methodology},
year={2022},
publisher={ACM New York, NY}
}