unveiled-the-red-hat/SEE-Few

Code for "SEE-Few: Seed, Expand and Entail for Few-shot Named Entity Recognition", accepted at COLING 2022.

SEE-Few

Code for "SEE-Few: Seed, Expand and Entail for Few-shot Named Entity Recognition", accepted at COLING 2022. For details of the model and experiments, please see our paper.

Architecture

Requirements

conda create --name see-few python=3.8
conda activate see-few
pip install -r requirements.txt

Data

To make the experimental results more convincing and credible, we randomly sample 5 different groups of training sets and validation sets for each K. All the datasets have been uploaded to datasets.

Train

bash scripts/run_conll.sh

Acknowledgements

We thank the authors for sharing their codes of Locate and Label, StructShot, TemplateNER and BaselineCode.

Contact

If you have any questions, please feel free to email yangzeng@seu.edu.cn.

Citation

@inproceedings{yang-etal-2022-see,
    title = "{SEE}-Few: Seed, Expand and Entail for Few-shot Named Entity Recognition",
    author = "Yang, Zeng  and
      Zhang, Linhai  and
      Zhou, Deyu",
    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-1.224",
    pages = "2540--2550",
}