BioMegatron: Larger Biomedical Domain Language Model

There has been an influx of biomedical domain-specific language models, showing language models pre-trained on biomedical text perform better on biomedical domain benchmarks than those trained on general domain text corpora such as Wikipedia and Books. Yet, most works do not study the factors affecting each domain language application deeply. Additionally, the study of model size on domain-specific models has been mostly missing. We empirically study and evaluate several factors that can affect performance on domain language applications, such as the sub-word vocabulary set, model size, pre-training corpus, and domain transfer. We show consistent improvements on benchmarks with our larger BioMegatron model trained on a larger domain corpus, contributing to our understanding of domain language model applications. We demonstrate noticeable improvements over the previous state-of-the-art (SOTA) on standard biomedical NLP benchmarks of named entity recognition, relation extraction, and question answering. Model checkpoints and code are available at [https://ngc.nvidia.com] and [https://github.com/NVIDIA/NeMo].

PDF Abstract EMNLP 2020 PDF EMNLP 2020 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Named Entity Recognition (NER) BC5CDR-chemical BioMegatron F1 92.9 # 12
Named Entity Recognition (NER) BC5CDR-disease BioMegatron F1 88.5 # 1
Relation Extraction ChemProt BioMegatron F1 77.0 # 7
Named Entity Recognition (NER) NCBI-disease BioMegatron BERT-cased F1 87.8 # 18

Methods


No methods listed for this paper. Add relevant methods here