Bioformer: an efficient transformer language model for biomedical text mining
This work addresses efficiency issues for researchers and practitioners in biomedical text mining, offering a more deployable model, though it is incremental as it builds on existing BERT adaptations.
The authors tackled the problem of computational expense in biomedical language models by developing Bioformer, a compact BERT model that reduces parameters by 60% with minimal accuracy loss (e.g., 0.1% less accurate than PubMedBERT) and achieves 2-3 times faster inference speeds.
Pretrained language models such as Bidirectional Encoder Representations from Transformers (BERT) have achieved state-of-the-art performance in natural language processing (NLP) tasks. Recently, BERT has been adapted to the biomedical domain. Despite the effectiveness, these models have hundreds of millions of parameters and are computationally expensive when applied to large-scale NLP applications. We hypothesized that the number of parameters of the original BERT can be dramatically reduced with minor impact on performance. In this study, we present Bioformer, a compact BERT model for biomedical text mining. We pretrained two Bioformer models (named Bioformer8L and Bioformer16L) which reduced the model size by 60% compared to BERTBase. Bioformer uses a biomedical vocabulary and was pre-trained from scratch on PubMed abstracts and PubMed Central full-text articles. We thoroughly evaluated the performance of Bioformer as well as existing biomedical BERT models including BioBERT and PubMedBERT on 15 benchmark datasets of four different biomedical NLP tasks: named entity recognition, relation extraction, question answering and document classification. The results show that with 60% fewer parameters, Bioformer16L is only 0.1% less accurate than PubMedBERT while Bioformer8L is 0.9% less accurate than PubMedBERT. Both Bioformer16L and Bioformer8L outperformed BioBERTBase-v1.1. In addition, Bioformer16L and Bioformer8L are 2-3 fold as fast as PubMedBERT/BioBERTBase-v1.1. Bioformer has been successfully deployed to PubTator Central providing gene annotations over 35 million PubMed abstracts and 5 million PubMed Central full-text articles. We make Bioformer publicly available via https://github.com/WGLab/bioformer, including pre-trained models, datasets, and instructions for downstream use.