ECPE-MM-R

Data and codes for the COLING2022 paper: A Multi-turn Machine Reading Comprehension Framework with Rethink Mechanism for Emotion-Cause Pair Extraction

If you use our codes or your research is related to our paper, please kindly cite our paper:

@inproceedings{zhou-etal-2022-multi-turn,
    title = "A Multi-turn Machine Reading Comprehension Framework with Rethink Mechanism for Emotion-Cause Pair Extraction",
    author = "Zhou, Changzhi  and
      Song, Dandan  and
      Xu, Jing  and
      Wu, Zhijing",
    booktitle = "Proceedings of the 29th International Conference on Computational Linguistics",
    month = oct,
    year = "2022",
    address = "Gyeongju, Republic of Korea",
    publisher = "International Committee on Computational Linguistics",
    url = "https://aclanthology.org/2022.coling-1.584",
    pages = "6726--6735",
}

Some code is based on BMRC, Rank-Emotion-Cause, and DeepInf.

Dependencies

• Python==3.7
• PyTorch==1.6
• Transformers from Hugging Face

The code has been tested on Ubuntu 18.04 using a single V100 GPU.

Quick Start

1. Clone or download this repo.
2. Download the pretrained parameters "bert-base-chinese/pytorch_model.bin" from this link. And then put the pytorch_model.bin to the folder pretrained_model/bert-base-chinese.
3. Run data preprocessing.
   • python3 dataProcess.py
   • python3 makeData_dual.py
4. Run our model MM-R.
   • sh run.sh