An imConvNet-based deep learning model for Chinese medical named entity recognition

Table 7 Performance comparison of different models on the yidu-s4k dataset

Model	Evaluation index (%)	Entity type						Comprehensive value
Model	Evaluation index (%)	Disease	Position	LabCheck	Check	Drug	Method	Comprehensive value
IDCNN-CRF	P(precision)	71.94	72.07	81.35	81.56	77.51	77.55	74.09
	R(recall)	70.88	78.83	76.67	76.25	72.38	78.35	75.77
	F1-score	71.41	75.30	78.94	78.81	74.86	77.95	74.92
BiLSTM-CRF	P(precision)	75.13	71.14	77.78	81.42	77.88	73.63	73.98
	R(recall)	72.16	80.06	76.36	78.93	72.93	76.29	76.75
	F1-score	73.62	75.34	77.06	80.16	75.32	74.94	75.34
imConvNet-CRF	P(precision)	72.45	74.79	77.06	72.26	81.06	76.77	74.84
	R(recall)	75.81	78.02	79.39	80.84	80.39	78.35	77.99
	F1-score	74.10	76.37	78.21	76.31	80.72	77.55	76.38
imConvNet-BiLSTM-CRF	P(precision)	74.53	76.19	77.78	81.60	78.84	81.77	76.77
	R(recall)	70.78	78.72	82.73	78.16	75.14	76.29	76.49
	F1-score	72.61	77.43	80.18	79.84	76.94	78.93	76.63
BERT-imConvNet-CRF	P(precision)	89.77	90.74	95.62	91.89	96.58	91.74	89.27
	R(recall)	89.38	86.95	94.70	97.14	97.35	90.97	87.19
	F1-score	89.57	88.81	95.16	94.44	96.96	91.35	88.22
BERT-imConvNet-BiLSTM CRF	P(precision)	92.36	92.44	97.32	95.45	98.42	95.48	91.30
	R(recall)	94.73	92.48	96.39	96.00	99.20	93.28	91.45
	F1-score	93.53	92.46	96.85	95.73	98.81	94.37	91.38

ISSN: 1472-6947