Temporal Knowledge Graph Completion using a Linear Temporal Regularizer and Multivector Embeddings

Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 2569–2578 June 6–11, 2021. ©2021 Association for Computational Linguistics 2569 Temporal Knowledge Graph Completion using a Linear Temporal Regularizer and Multivector Embeddings Chengjin Xu University of Bonn / Germany xuc@iai.uni-bonn.de Yung-Yu Chen University of Bonn / Germany s6ynchen@uni-bonn.de Mojtaba Nayyeri University of Bonn / Germany nayyeri@iai.uni-bonn.de Jens Lehmann University of Bonn / Germany Fraunhofer IAIS/ Germany jens.lehmann@iais.fraunhofer.de Abstract Representation learning approaches for knowl- edge graphs have been mostly designed for static data. However, many knowledge graphs involve evolving data, e.g., the fact (The Presi- dent of the United States is Barack Obama) is valid only from 2009 to 2017. This introduces important challenges for knowledge represen- tation learning since the knowledge graphs change over time. In this paper, we present a novel time-aware knowledge graph em- bebdding approach, TeLM, which performs 4th-order tensor factorization of a Temporal knowledge graph using a Linear temporal reg- ularizer and Multivector embeddings. More- over, we investigate the effect of the temporal dataset’s time granularity on temporal knowl- edge graph completion. Experimental results demonstrate that our proposed models trained with the linear temporal regularizer achieve the state-of-the-art performances on link pre- diction over four well-established temporal knowledge graph completion benchmarks. 1 Introduction Numerous large-scale knowledge graphs (KGs) including DBpedia (Auer et al., 2007), Free- Base (Bollacker et al., 2008) and WordNet (Miller, 1995) have been established in recent years. Such KGs abstract knowledge from the real world into a complex network graph consisting of billions of triples. Each triple is denoted as (s, r, o), where s is the subject entity, o is the object entity, and r is the relation between the entities. Knowledge graph completion is one of the main challenges in the KG field since most KGs are in- complete. To tackle this problem, knowledge graph embedding (KGE) approaches embed entities and relations into a low-dimensional embedding space and measure the plausibility of triples by inputting embeddings of the entities and their relation to a score function (Wang et al., 2017). For instance, ComplEx (Trouillon et al., 2016) has been proven to be a highly effective KGE model, where entities and relations are represented as complex embed- dings, and the score of a triple (s, r, o) is computed with the asymmetric Hermitian dot product. Some KGs involve temporal facts, e.g., the triple (Barack Obama, presidentOf, USA) is only valid in a specific time period [2009, 2017]. Temporal KGs like Wikidata (Erxleben et al., 2014), YAGO3 (Mahdisoltani et al., 2013) and ICEWS (Lautenschlager et al., 2015) incorporate time information into triples. Triples attached with time information are represented as quadruples, shaped like (s, r, o, T), where T denotes the times- tamp. Traditional KGE models disregard time in- formation, leading to an ineffectiveness of perform- ing link prediction on TKGs involving temporary relations, e.g., (?, presidentOf, USA, 2010). Re- cent researches show that the temporal knowledge graph embedding (TKGE) models, which encode …

Content truncated to 500 words