notesum.ai

Published at December 6

eXpath: Explaining Knowledge Graph Link Prediction with Ontological Closed Path Rules

cs.AI

cs.DB

cs.IR

cs.LG

Released Date: December 6, 2024

Authors: Ye Sun¹, Lei Shi¹, Yongxin Tong¹

Aff.: ¹School of Computer Science, Beihang University, Beijing, China

Arxiv: http://arxiv.org/pdf/2412.04846v1

Refer to caption

Max Exp. Size	Method	Complex				Conve				TransE				AVG
Max Exp. Size	Method	FB15k	FB15k-237	WN18	WN18RR	FB15k	FB15k-237	WN18	WN18RR	FB15k	FB15k-237	WN18	WN18RR	AVG
single -fact exp.	Criage (Pezeshkpour et al., 2019)	.087	.105	.080	.203	.153	.162	.270	.256	—	—	—	—	.165
	DP (Zhang et al., 2019b)	.529	.315	.799	.758	.246	.162	.794	.829	.304	.326	.910	.709	.557
	Kelpie (Rossi et al., 2022)	.576	.395	.578	.593	.229	.222	.567	.667	.261	.281	.792	.779	.495
	KGEAttack (Betz et al., 2022)	.547	.290	.829	.764	.237	.212	.929	.915	.365	.213	.938	.779	.585
	eXpath	.512	.395	.834	.797	.271	.343	.929	.891	.313	.337	.938	.767	.611
	Criage+eXPath	.523	.411	.839	.819	.322	.404	.936	.891	—	—	—	—	.643 (+290%)
	DP+eXpath	.570	.500	.859	.813	.331	.414	.936	.946	.374	.438	.944	.826	.663 (+19%)
	Kelpie+eXPath	.657	.540	.859	.835	.364	.424	.929	.915	.417	.427	.944	.872	.682 (+38%)
	KGEA.+eXpath	.576	.452	.859	.802	.322	.384	.929	.946	.417	.360	.938	.872	.655 (+12%)
four -fact exp.	Kelpie	.767	.581	.829	.940	.534	.303	.816	.946	.374	.427	.868	.907	.691
	eXpath	.802	.661	.920	.951	.542	.566	.957	.984	.539	.573	.965	.965	.785
	Kelpie+eXpath	.831	.742	.935	.989	.653	.596	.965	.984	.609	.674	.965	.965	.826