Answers are grounded in this paper and cite their sources.
Experimental evaluation demonstrates enhanced reasoning accuracy across three public graph datasets, highlighting the benefit of constrained global temporal graphs.
The objective of temporal knowledge graph reasoning (TKGR) is to infer the absent component of the quadruple based on the historical temporal knowledge graphs (TKGs). Some of the current TKGR models adequately capture the local information in historical TKGs while others utilize historical matrices to extract recurrent historical facts. Additionally some models consider the adverse effects of repeated historical facts on TKGR tasks, and others extract global information of historical TKGs. However, these models ignore that a large number of historical TKGs may have a negative impact when mining global information. This paper constructs a TKG reasoning model based on Global temporal graph and History matrix, abbreviated as TRGH model. The model constructs a global temporal graph, fully mining the local and global information in the historical TKGs, and constructs a history matrix to provide global constraints for the decoder, reducing the negative impact of a large number of historical TKGs. Numerous experiments have been conducted on three public datasets. The experimental results show that the proposed TRGH model is superior to most of the advanced models, and demonstrates its efficacy in mitigating the adverse effects of several previous TKGs in the extraction of global information.
No takes yet. Share an insight, caveat, or question.
Liu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: