TeAST: Temporal Knowledge Graph Embedding via Archimedean Spiral Timeline

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 15460–15474 July 9-14, 2023 ©2023 Association for Computational Linguistics TeAST: Temporal Knowledge Graph Embedding via Archimedean Spiral Timeline Jiang Li1,2 , Xiangdong Su1,2 ∗, Guanglai Gao1,2 1 College of Computer Science, Inner Mongolia University, Hohhot, China 2 National & Local Joint Engineering Research Center of Intelligent Information Processing Technology for Mongolian, Hohhot, China lijiangimu@gmail.com, cssxd@imu.edu.cn, csggl@imu.edu.cn Abstract Temporal knowledge graph embedding (TKGE) models are commonly utilized to infer the missing facts and facilitate reasoning and decision-making in temporal knowledge graph based systems. However, existing methods fuse temporal information into entities, potentially leading to the evolution of entity information and 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 Temporal knowledge graph embeddings via Archimedean Spiral Timeline (TeAST), which maps relations onto the corresponding Archimedean spiral timeline and transforms the quadruples completion to 3th-order tensor com- pletion problem. Specifically, the Archimedean spiral timeline ensures that relations that occur simultaneously are placed 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. Our code is available at https://github.com/ IMU-MachineLearningSXD/TeAST. 1 Introduction Knowledge graph (KG) expresses the relations of real-world entities and allows for reasoning new facts, which enables a wide range of applications in natural language processing (Chen et al., 2019; Junior et al., 2020; Hu et al., 2021). It stores a vast amount of knowledge in the form of triplets. These triplets are typically denoted as (s, r, o), ∗Corresponding Author Figure 1: A brief illustration of mapping relations to Archimedean spiral timeline. Three facts are (Guido van Rossum, r1, Google, 2010-12), (Guido van Rossum, r2, Google, 2012-12) and (Messi, r3, FCB, 2010-12). where s, r and o represent the subject, the rela- tion, and the object. Since knowledge changes over time, researchers introduced timestamps into knowledge graphs to create temporal knowledge graphs (TKGs). In TKGs, each knowledge fact is represented as a quadruple (s, r, o, τ), where τ denotes the timestamp at which the fact was true. This allows for more precise representation and querying of information in knowledge graphs, en- abling applications that require an understanding of the evolution of knowledge over time. Given the inherent incompleteness of most KGs and TKGs, knowledge graph embedding (KGE) and temporal knowledge graph embedding (TKGE) have been widely investigated to infer the missing facts using the existing ones. In particular, TKGE has gained significant attention for its ability to represent and analyze knowledge over time. This work focuses on TKGE. With the advancement of deep learning, re- …

Content truncated to 500 words