Skip to main content

Table 3 Comparison metrics of biomedically related pre-trained BERT-based models

From: Precision information extraction for rare disease epidemiology at scale

Model

Overall token-level precision

Overall entity-level precision

BioBERT large cased v1.1

0.834

0.720

BioBERT base cased v1.2

0.825

0.698

PubMedBERT base

0.824

0.713

PubMedBERT + PMC base

0.824

0.715

BlueBERT base

0.829

0.722

BlueBERT large

0.818

0.690

  1. The highest scores in each category are bolded. The second highest are underlined