SapBERT: Self-alignment pretraining for BERT

This repo holds code for the SapBERT model presented in our NAACL 2021 paper: Self-Alignment Pretraining for Biomedical Entity Representations [arxiv]; and our ACL 2021 paper: Learning Domain-Specialised Representations for Cross-Lingual Biomedical Entity Linking [PDF].

Huggingface Models

[SapBERT]

Standard SapBERT as described in [Liu et al., NAACL 2021]. Trained with UMLS 2020AA (English only), using microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext as the base model. Use [CLS] (before pooler) as the representation of the input.

[SapBERT-XLMR]

Cross-lingual SapBERT as described in [Liu et al., ACL 2021]. Trained with UMLS 2020AB (all languages), using xlm-roberta-base as the base model. Use [CLS] (before pooler) as the representation of the input.

[SapBERT-mean-token]

Same as the standard SapBERT but trained with mean-pooling instead of [CLS] representations.

Environment

The code is tested with python 3.8, torch 1.7.0 and huggingface transformers 4.4.2. Please view requirements.txt for more details.

Train SapBERT

Prepare training data as insrtructed in data/generate_pretraining_data.ipynb.

Run:

cd umls_pretraining
./pretrain.sh 0,1 

where 0,1 specifies the GPU devices.

Evaluate SapBERT

Please view evaluation/README.md for details.

Citations

@article{liu2021self,
	title={Self-Alignment Pretraining for Biomedical Entity Representations},
	author={Liu, Fangyu and Shareghi, Ehsan and Meng, Zaiqiao and Basaldella, Marco and Collier, Nigel},
	journal={arXiv preprint arXiv:2010.11784},
	year={2020}
}

Acknowledgement

Parts of the code are modified from BioSyn. We appreciate the authors for making BioSyn open-sourced.

License

SapBERT is MIT licensed. See the LICENSE file for details.