TeAST: Temporal Knowledge Graph Embedding via Archimedean Spiral Timeline

Anonymous

Abstract: Temporal knowledge graph embedding (TKGE) models are often utilized to infer the missing facts and facilitate reasoning and decision-making in temporal knowledge graph based systems. However, existing TKGE methods fuse temporal information into entities leading to the potential evolution of entity information, thus limiting the link prediction performance of TKG. Meanwhile, current TKGE models often lack the ability to simultaneously model important relation patterns and provide interpretability, which hinders their effectiveness and potential applications. To address these limitations, we propose a novel TKGE model which encodes \textbf{T}emporal knowledge graph \textbf{e}mbeddings via \textbf{A}rchimedean \textbf{S}piral \textbf{T}imeline (TeAST), which maps relations onto the corresponding Archimedean spiral timeline and transforms the quadruples completion to 3th-order tensor completion problem. Specifically, the Archimedean spiral timeline ensures that relations at the same time to be on the same timeline and all relations evolve over time. Meanwhile, we present a novel temporal spiral regularizer to make the spiral timeline orderly. In addition, we provide mathematical proofs to demonstrate the ability of TeAST to encode various relation patterns. Experimental results show that our proposed model significantly outperforms existing TKGE methods. The code of our paper is available online: \url{https://anonymous.4open.science/r/teast-D4D4/}.

Paper Type: long

Research Area: Efficient Methods for NLP