A Knowledge Graph Entity Disambiguation Method Based on Entity-Relationship Embedding and Graph Structure Embedding

Table 3

Disambiguation results of EDEGE and baseline methods with different knowledge graphs.

Data set	Knowledge graph	Method	F1	Precision	Recall
AIDA-CoNLL	DBpedia	PBoH	0.694	0.766	0.635
		DoSeR	0.708	0.791	0.641
		NeuPL	0.719	0.803	0.651
		GNED	0.729	0.809	0.663
		NCEL	0.717	0.806	0.656
		E2ENEL	0.704	0.805	0.658
		BOOTLEG	0.721	0.803	0.651
		EDEGE	0.801	0.836	0.768
AIDA-B	DBpedia	PBoH	0.668	0.698	0.641
		DoSeR	0.682	0.714	0.652
		NeuPL	0.701	0.737	0.669
		GNED	0.715	0.751	0.683
		NCEL	0.709	0.743	0.679
		E2ENEL	0.711	0.739	0.681
		BOOTLEG	0.708	0.744	0.676
		EDEGE	0.777	0.812	0.744
MSNBC	DBpedia	PBoH	0.677	0.705	0.652
		DoSeR	0.694	0.724	0.667
		NeuPL	0.705	0.698	0.713
		GNED	0.706	0.727	0.686
		NCEL	0.703	0.713	0.676
		E2ENEL	0.698	0.708	0.681
		BOOTLEG	0.702	0.715	0.679
		EDEGE	0.779	0.823	0.739
AQUAINT	DBpedia	PBoH	0.673	0.699	0.648
		DoSeR	0.686	0.718	0.657
		NeuPL	0.696	0.732	0.663
		GNED	0.706	0.728	0.686
		NCEL	0.693	0.724	0.679
		E2ENEL	0.702	0.713	0.681
		BOOTLEG	0.689	0.709	0.673
		EDEGE	0.792	0.851	0.741
ACE2004	DBpedia	PBoH	0.661	0.684	0.639
		DoSeR	0.680	0.705	0.657
		NeuPL	0.705	0.734	0.679
		GNED	0.732	0.762	0.705
		NCEL	0.724	0.753	0.701
		E2ENEL	0.718	0.741	0.698
		BOOTLEG	0.726	0.729	0.703
		EDEGE	0.824	0.832	0.817

To distinguish the proposed EDEGE method from baseline methods, the maximum value is bold in each column.