Entity recognition in Chinese clinical text using attention-based CNN-LSTM-CRF

Table 2 Performances of different methods on the two datasets: CCKS2017_CNER and ICRC_CNER

Dataset	Method	Strict (%)			Relaxed (%)
Dataset	Method	Precision	Recall	F1-score	Precision	Recall	F1-score
CCKS2017_CNER	CRF	91.22	88.20	89.69	95.73	92.57	94.13
	LSTM-CRF	90.68	89.67	90.17	95.18	94.12	94.65
	Our Method	90.73	90.49	90.61	94.84	94.59	94.71
ICRC_CNER	CRF	81.84	78.86	80.32	93.75	90.34	92.01
	CRF	83.42	79.90	81.62	94.02	90.05	92.00
	LSTM-CRF	83.55	82.26	82.90	93.80	92.35	93.07
	LSTM-CRF	82.71	83.30	83.00	92.77	93.42	93.09
	Our Method	82.96	82.60	82.78	93.30	92.90	93.10
	Our Method	82.66	83.99	83.32	92.57	94.07	93.31

Table 2 shows the performances of different methods on CCKS2017_CNER and ICRC_CNER, where the highest measures are in bold (the following sections also use the same way to denote the highest measures)

ISSN: 1472-6947