wyxstriker / ReweightingDisfluency

Implementation of COLING 2022 paper "Adaptive Unsupervised Self-training for Disfluency Detection"

ReweightingDisfluency

This is the PyTorch implementation of the

• COLING2022 paper Adaptive Unsupervised Self-training for Disfluency Detection

All the code and model are released. Thank you for your patience!

About Model

We release our self-supervised model trained by pseudo data and grammar check model. Please download it in the following link, and put model in "./ckpt/teacher" and "./ckpt/judge " folder.

• teacher_model in "./ckpt/teacher"

• judge_model in "./ckpt/judge"

You need to put your data and model in the parallel folder of this repo:

    - ckpt/
        - electra_en_base
            - config.json
            - pytorch_model.bin
            - vocab.txt
        - teacher
            - pytorch_model.bin
        - judge
            - pytorch_model.bin
    - self_training/
        - run_data/
            - 500/
                - unlabel.tsv
                - dev.tsv
                - test.tsv
        - run_model/
    - src/
        - model.py
        ...
    - run.sh

About data

Due to copyright issues, we do not have the right to distribute the SWBD dataset and can purchase it for your own use.

Requirements

• transformers==4.7.0
• pytorch==1.9
• numpy
• tensorboardX

How to use

The file path and training details can be set in the script run.sh

nohup sh run.sh > log_run 2>&1 &

Citation

If you find this project useful for your research, please consider citing the following paper:

@inproceedings{wang2022adaptive,
  title={Adaptive Unsupervised Self-training for Disfluency Detection},
  author={Wang, Zhongyuan and Wang, Yixuan and Wang, Shaolei and Che, Wanxiang},
  booktitle={Proceedings of the 29th International Conference on Computational Linguistics},
  pages={7209--7218},
  year={2022}
}

Contact

If you have any question about this code, feel free to open an issue or contact [e-mail].