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

Table 7

Comparison of different methods on CCKS2019.

Method	Precision	Recall	F1 score
BERT-IDCNN-MHA-CRF (Liang et al. [37])	82.63	82.23	82.43
BiLSTM-CRF (baseline)	81.49	80.52	81.00
IDCNN-CRF (baseline)	80.56	81.47	81.01
Ours	83.31	82.44	82.87