Research on Named Entity Recognition Method of Metro On-Board Equipment Based on Multiheaded Self-Attention Mechanism and CNN-BiLSTM-CRF

Table 2

Recognition effect of different named entity recognition models on phenomenon.

Models	B-phenomenon			M-phenomenon			E-phenomenon
	P (%)	R (%)	F1 (%)	P (%)	R (%)	F1 (%)	P (%)	R (%)	F1 (%)
HMM	55.58	79.17	65.52	74.90	86.61	80.33	62.12	85.42	71.93
CRF	74.42	66.67	70.33	86.49	78.57	82.34	81.40	72.92	76.92
BiLSTM	83.78	64.58	72.94	84.92	85.49	85.21	78.72	77.08	77.89
BiLSTM-CRF	81.25	66.72	73.27	88.67	80.67	84.48	84.85	75.09	79.67
BiLSTM-CNN-CRF	84.85	66.67	74.67	95.59	78.33	86.11	91.43	76.19	83.12
BiLSTM-CNN-MHA-CRF	78.95	71.43	75.00	93.67	82.22	87.57	87.18	80.95	83.95