GradLRE: Gradient Imitation Reinforcement Learning for Low resource Relation Extraction.

This project provides tools for "GradLRE: Gradient Imitation Reinforcement Learning for Low resource Relation Extraction." in EMNLP as a long paper.

Details about low resource relation are in the paper and the implementation is based on the PyTorch library.

Installation

For training, a GPU is recommended to accelerate the training speed.

PyTroch

The code is based on PyTorch 1.6+. You can find tutorials here.

Dependencies

The code is written in Python 3.7. Its dependencies are summarized in the file requirements.txt.

torch==1.6.0
numpy==1.18.5
scikit_learn==0.23.2
transformers==3.5.1
tqdm==4.48.2

You can install these dependencies like this:

pip3 install -r requirements.txt

Usage

Run the full model on SemEval dataset with default hyperparameter settings

python3 src/train.py

If you need data augmentation to generate unlabeled data in low resource scenarios, please run with the following parameter

python3 src/train.py --use_aug True

Data

Format

Each dataset is a folder under the ./data folder:

./data
└── SemEval
    ├── train_sentence.json
    ├── train_label_id.json
    ├── dev_sentence.json
    ├── dev_label_id.json
    ├── test_sentence.json
    └── test_label_id.json

Download

SemEval: SemEval 2010 Task 8 data (included in data/SemEval)
TACRED: The TAC Relation Extraction Dataset (download)

Then use the scripts from data/data_prepare.py to further preprocess the data. For SemEval, the script split the original training data into two sets. For TACRED, the script first perform some preprocessing to ensure the same format as SemEval.

Acknowledgements

https://github.com/huggingface/transformers

https://github.com/INK-USC/DualRE

Contact

If you have any problem about our code, feel free to contact: hxm19@mails.tsinghua.edu.cn

Reference

If the code is used in your research, hope you can cite our paper as follows:

@inproceedings{hu2021gradient,
  abbr = {EMNLP},
  title = {Gradient Imitation Reinforcement Learning for Low Resource Relation Extraction},
  author = {Hu, Xuming and Zhang, Chenwei and Yang, Yawen and Li, Xiaohe and Lin, Li and Wen, Lijie and Yu, Philip S.},
  booktitle = {Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing},
  year = {2021},
  pdf = {https://arxiv.org/pdf/2109.06415.pdf},
  code = {https://github.com/THU-BPM/GradLRE}
}

THU-BPM / GradLRE