ContProto

This repository contains the code for the ACL 2023 paper "Improving Self-training for Cross-lingual Named Entity Recognition with Contrastive and Prototype Learning"

Requirements

• torch>=1.6
• pytorch-lightning>=0.9.0
• tokenizers
• transformers

To install the dependencies, run:

pip install -r requirements.txt

Our experiments are run on a single Nvidia V100 32GB GPU.

Data Format

Before running the experiments, you will need to convert the NER dataset from CoNLL format to span-based format by running the following scripts:

cd data
bash data_preprocess.sh

We have provided the processed German dataset from CoNLL as an example in data/conll03_de.

Generating Pseudo-labeled Data

To generate the pseudo-labeled data, run the following script:

bash run_generate_pseudo.sh de conll03

For example, the generated German pseudo labels will be stored in train_logs/conll03/genpseudo_de_0/.

Training

To start training of ContProto, run the following script:

bash run_contpro_auto-margin.sh de conll03

Citation

If you find this repository useful, please cite our paper:

@inproceedings{zhou-etal-2023-improving,
    title = "Improving Self-training for Cross-lingual Named Entity Recognition with Contrastive and Prototype Learning",
    author = "Zhou, Ran  and
      Li, Xin  and
      Bing, Lidong  and
      Cambria, Erik  and
      Miao, Chunyan",
    booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    year = "2023",
}