Combining BERT Model with Semi-Supervised Incremental Learning for Heterogeneous Knowledge Fusion of High-Speed Railway On-Board System

Table 2

Experimental results under different EM models.

	Entity representation	Models	Marco-P	Marco-R	Marco-F
I	Embedding-based models	ABCNN-3	0.8782	0.8762	0.8765
		LSTM-Siamese	0.8553	0.8219	0.8374
		BIMPM	0.9078	0.8998	0.8985
		ESIM	0.9127	0.9093	0.9086
II	BERT-based models	ALBERT	0.9553	0.9511	0.9529
		ERNIE	0.9729	0.9711	0.9720
III		AD--SSI	0.9887	0.9883	0.9885
		AD--SSI	0.9859	0.9856	0.9857
		AD--SSI	0.9871	0.9863	0.9867