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

Table 3

Recognition effect of different named entity recognition models on location.

Models	B-location			M-location			E-location
	P (%)	R (%)	F1 (%)	P (%)	R (%)	F1 (%)	P (%)	R (%)	F1 (%)
HMM	92.92	91.67	92.29	91.26	92.57	91.91	91.18	86.11	88.57
CRF	96.88	86.11	91.18	95.28	86.43	90.64	93.75	83.33	88.24
BiLSTM	96.43	75.00	84.37	92.81	92.14	92.47	91.43	88.89	90.14
BiLSTM-CRF	94.87	90.24	92.50	91.30	90.74	91.02	92.31	87.80	90.00
BiLSTM-CNN-CRF	97.30	87.80	92.31	92.31	87.80	90.00	93.67	91.36	92.50
BiLSTM-CNN-MHA-CRF	97.37	90.24	93.67	94.27	91.36	92.79	94.87	90.24	92.50