Chinese medical entity recognition based on the dual-branch TENER model

Table 5 Overall results of data augmentation, where Rt stands for R-Transformer, TE stands for TENER

	CCKS2019			ERTCMM			CMeEE			CEMR
Model	P	R	F1	P	R	F1	P	R	F1	P	R	F1
BERT-Rt-CRF	82.42	83.23	82.82	61.42	72.44	66.48	63.45	65.27	64.34	63.04	62.09	62.56
TENER-LSTM-CRF	84.99	87.12	86.02	65.65	74.19	69.66	64.07	67.29	65.64	66.54	64.92	65.72
TENER-Rt-CRF	83.14	81.71	81.20	67.35	75.33	71.12	64.53	63.08	63.79	67.22	66.82	67.03
RoBERTa-TE-CRF	84.11	87.24	85.65	67.25	73.22	70.11	62.74	60.11	61.39	66.23	67.06	66.60
RoBERTa-Rt-CRF	85.61	86.24	85.92	67.68	73.45	70.45	62.52	58.02	60.18	67.38	68.65	68.01
ELECTRA-Rt-CRF	86.18	86.23	86.20	68.19	76.44	72.08	64.71	63.32	64.00	69.25	66.03	67.60
Ours	85.70	87.18	86.41	68.34	81.56	74.37	67.45	69.37	68.39	69.32	71.04	70.16

ISSN: 1472-6947