Home

Awesome

BioBERT Pre-trained Weights

This repository provides pre-trained weights of BioBERT, a language representation model for biomedical domain, especially designed for biomedical text mining tasks such as biomedical named entity recognition, relation extraction, question answering, etc. Please refer to our paper BioBERT: a pre-trained biomedical language representation model for biomedical text mining for more details.

Downloading pre-trained weights

Go to releases section of this repository or click links below to download pre-trained weights of BioBERT. We provide three combinations of pre-trained weights: BioBERT (+ PubMed), BioBERT (+ PMC), and BioBERT (+ PubMed + PMC). Pre-training was based on the original BERT code provided by Google, and training details are described in our paper. Currently available versions of pre-trained weights are as follows:

Make sure to specify the versions of pre-trained weights used in your works. If you have difficulty choosing which one to use, we recommend using BioBERT-Base v1.1 (+ PubMed 1M) or BioBERT-Large v1.1 (+ PubMed 1M) depending on your GPU resources. Note that for BioBERT-Base, we are using WordPiece vocabulary (vocab.txt) provided by Google as any new words in biomedical corpus can be represented with subwords (for instance, Leukemia => Leu + ##ke + ##mia). More details are in the closed issue #1.

Pre-training corpus

We do not provide pre-processed version of each corpus. However, each pre-training corpus could be found in the following links:

Estimated size of each corpus is 4.5 billion words for PubMed Abstracts1 + PubMed Abstracts2, and 13.5 billion words for PubMed Central Full Texts.

Fine-tuning BioBERT

To fine-tunine BioBERT on biomedical text mining tasks using provided pre-trained weights, refer to the DMIS GitHub repository for BioBERT.

Citation

@article{10.1093/bioinformatics/btz682,
    author = {Lee, Jinhyuk and Yoon, Wonjin and Kim, Sungdong and Kim, Donghyeon and Kim, Sunkyu and So, Chan Ho and Kang, Jaewoo},
    title = "{BioBERT: a pre-trained biomedical language representation model for biomedical text mining}",
    journal = {Bioinformatics},
    year = {2019},
    month = {09},
    issn = {1367-4803},
    doi = {10.1093/bioinformatics/btz682},
    url = {https://doi.org/10.1093/bioinformatics/btz682},
}

Contact information

For help or issues using pre-trained weights of BioBERT, please submit a GitHub issue. Please contact Jinhyuk Lee (lee.jnhk@gmail.com), or Sungdong Kim (sungdong.kim@navercorp.com) for communication related to pre-trained weights of BioBERT.