Electronic Medical Record Entity Recognition via Machine Reading Comprehension and Biaffine

Table 4

The result for the entity recognition task (Precision, Recall, and F1).

Methods	Evaluation	Entity categories
		BODY	CHECK	DISEASE	SIGNS	TREATMENT	OVERALL
BiLSTM-CRF	P	85.87	94.51	85.51	90.03	71.91	89.15
	R	83.16	95.3	67.05	98.53	74.42	89.94
	F1	84.49	94.9	75.16	94.09	73.14	89.54
BERT-BiLSTM-CRF	P	87.82	89.50	91.11	89.40	63.16	86.43
	R	85.16	93.83	73.21	97.92	69.23	88.32
	F1	86.47	91.61	81.19	93.47	66.06	87.36
BERT-Softmax	P	86.38	87.92	88.70	85.43	62.48	84.47
	R	83.68	86.39	72.09	93.71	67.13	86.04
	F1	85.01	87.15	79.54	89.38	64.72	85.25
BERT-MRC-Biaffine	P	91.68	96.84	94.20	95.33	75.93	91.93
	R	89.06	93.91	81.32	96.40	86.82	93.68
	F1	90.35	95.35	87.29	95.86	81.01	92.80