RotateQVS: Representing Temporal Information as Rotations in Quaternion Vector Space for Temporal Knowledge Graph Completion

Temporal factors are tied to the growth of facts in realistic applications,\nsuch as the progress of diseases and the development of political situation,\ntherefore, research on Temporal Knowledge Graph (TKG) attracks much attention.\nIn TKG, relation patterns inherent with temporality are required to be studied\nfor representation learning and reasoning across temporal facts. However,\nexisting methods can hardly model temporal relation patterns, nor can capture\nthe intrinsic connections between relations when evolving over time, lacking of\ninterpretability. In this paper, we propose a novel temporal modeling method\nwhich represents temporal entities as Rotations in Quaternion Vector Space\n(RotateQVS) and relations as complex vectors in Hamilton's quaternion space. We\ndemonstrate our method can model key patterns of relations in TKG, such as\nsymmetry, asymmetry, inverse, and can further capture time-evolved relations by\ntheory. Empirically, we show that our method can boost the performance of link\nprediction tasks over four temporal knowledge graph benchmarks.\n