Research on Named Entity Recognition of Electronic Medical Records Based on RoBERTa and Radical-Level Feature

Table 2

Comparison of three pretraining models on CCKS2017.

Model	Precision	Recall	F1 score
BERT+CRF	89.78	91.64	90.70
ALBERT+CRF	89.12	91.38	90.24
RoBERTa-wwm-ext+CRF	92.62	90.35	91.47
RoBERTa-wwm-ext-large+CRF	92.79	91.34	92.06