Skip to main navigation Skip to search Skip to main content

Each Snapshot to Each Space: Space Adaptation for Temporal Knowledge Graph Completion

  • Yancong Li
  • , Xiaoming Zhang*
  • , Bo Zhang
  • , Haiying Ren
  • *Corresponding author for this work
  • Beihang University

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Temporal knowledge graphs (TKGs) organize and manage the dynamic relations between entities over time. Inferring missing knowledge in TKGs, known as temporal knowledge graph completion (TKGC), has become an important research topic. Previous models handle all facts with different timestamps in an identical latent space, even though the semantic space of the TKG changes over time. Therefore, they are not effective to reflect the temporality of knowledge. To effectively learn the time-aware information of TKGs, different latent spaces are adapted for temporal snapshots at different timestamps, which yields a novel model, i.e., Space Adaptation Network (SANe). Specifically, we extend convolutional neural networks (CNN) to map the facts with different timestamps into different latent spaces, which can effectively reflect the dynamic variation of knowledge. Meanwhile, a time-aware parameter generator is designed to explore the overlap of latent spaces, which endows CNN with specific parameters in term of the context of timestamps. Therefore, knowledge in adjacent time intervals is efficiently shared to boost the performance of TKGC, which can learn the validity of knowledge over a period of time. Extensive experiments demonstrate that SANe achieves state-of-the-art performance on four well-established benchmark datasets for temporal knowledge graph completion.

Original languageEnglish
Title of host publicationThe Semantic Web – ISWC 2022 - 21st International Semantic Web Conference, Proceedings
EditorsUlrike Sattler, Aidan Hogan, Maria Keet, Valentina Presutti, João Paulo A. Almeida, Hideaki Takeda, Pierre Monnin, Giuseppe Pirrò, Claudia d’Amato
PublisherSpringer Science and Business Media Deutschland GmbH
Pages248-266
Number of pages19
ISBN (Print)9783031194320
DOIs
StatePublished - 2022
Event21st International Semantic Web Conference, ISWC 2022 - Virtual, Online
Duration: 23 Oct 202227 Oct 2022

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume13489 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference21st International Semantic Web Conference, ISWC 2022
CityVirtual, Online
Period23/10/2227/10/22

Keywords

  • Parameter generation
  • Space adaptation
  • Temporal knowledge graph
  • Temporal knowledge graph completion

Fingerprint

Dive into the research topics of 'Each Snapshot to Each Space: Space Adaptation for Temporal Knowledge Graph Completion'. Together they form a unique fingerprint.

Cite this