StRE

Implementation of the paper StRE: Self Attentive Edit Quality Prediction in Wikipedia

Organization of the folder

1. Preprocessing - preprocess the raw data files to required format as described in the paper.
2. Baselines - The ORES++ models that are described in the paper are implemented here.
3. Category experiments - The character, word, both, attention modules are implemented. Transfer learning is also implemented.
4. Big dataset - The same experiments as in category but scaled up for the big dataset. The retraining module is modified to handle for the files in a folder.

Baselines

1. The ORES++ baseline presented in the paper is given in the Baselines folder
2. Interrank and ORES baselines are taken from the repository lca4/interrank.

If you find this code useful in your research then please cite

To Run

Please look into the individual folders for running the codes.

If you find this code useful in your research then please cite

@inproceedings{soumya2019stre,
  title={StRE: Self Attentive Edit Quality Prediction in Wikipedia},
  author={Sarkar, Soumya and Guda, Bhanu Prakash Reddy and Sikdar, Sandipan and Mukherjee, Animesh},
  booktitle={ACL},
  year={2019}
}