Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Outdated Fact Detection in Knowledge Graphs

Huiling Tu, Shuo Yu, Vidya Saikrishna, Feng Xia, Karin Verspoor

TL;DR

This work tackles outdated fact detection in knowledge graphs by introducing DEAN, a deep learning framework that jointly learns entity and relation representations through a Fact Attention Module and a Contrastive R2N Graph Module. It initializes features with TransE, applies graph attention to capture local structure, and employs a contrastive, relation-aware objective on a predefined R2N graph to better differentiate outdated facts from their non-outdated counterparts. Empirical results across six datasets show DEAN outperforms a range of KG embedding baselines, especially when relation diversity is high, underscoring the value of implicit structural information and relational contrastive learning for KG quality control. The approach advances automatic outdated fact detection, with future work pointing toward handling entity changes and adapting to datasets with fewer relation types.

Abstract

Knowledge graphs (KGs) have garnered significant attention for their vast potential across diverse domains. However, the issue of outdated facts poses a challenge to KGs, affecting their overall quality as real-world information evolves. Existing solutions for outdated fact detection often rely on manual recognition. In response, this paper presents DEAN (Deep outdatEd fAct detectioN), a novel deep learning-based framework designed to identify outdated facts within KGs. DEAN distinguishes itself by capturing implicit structural information among facts through comprehensive modeling of both entities and relations. To effectively uncover latent out-of-date information, DEAN employs a contrastive approach based on a pre-defined Relations-to-Nodes (R2N) graph, weighted by the number of entities. Experimental results demonstrate the effectiveness and superiority of DEAN over state-of-the-art baseline methods.

Deep Outdated Fact Detection in Knowledge Graphs

TL;DR

This work tackles outdated fact detection in knowledge graphs by introducing DEAN, a deep learning framework that jointly learns entity and relation representations through a Fact Attention Module and a Contrastive R2N Graph Module. It initializes features with TransE, applies graph attention to capture local structure, and employs a contrastive, relation-aware objective on a predefined R2N graph to better differentiate outdated facts from their non-outdated counterparts. Empirical results across six datasets show DEAN outperforms a range of KG embedding baselines, especially when relation diversity is high, underscoring the value of implicit structural information and relational contrastive learning for KG quality control. The approach advances automatic outdated fact detection, with future work pointing toward handling entity changes and adapting to datasets with fewer relation types.

Abstract

Knowledge graphs (KGs) have garnered significant attention for their vast potential across diverse domains. However, the issue of outdated facts poses a challenge to KGs, affecting their overall quality as real-world information evolves. Existing solutions for outdated fact detection often rely on manual recognition. In response, this paper presents DEAN (Deep outdatEd fAct detectioN), a novel deep learning-based framework designed to identify outdated facts within KGs. DEAN distinguishes itself by capturing implicit structural information among facts through comprehensive modeling of both entities and relations. To effectively uncover latent out-of-date information, DEAN employs a contrastive approach based on a pre-defined Relations-to-Nodes (R2N) graph, weighted by the number of entities. Experimental results demonstrate the effectiveness and superiority of DEAN over state-of-the-art baseline methods.
Paper Structure (19 sections, 13 equations, 6 figures, 3 tables, 2 algorithms)

This paper contains 19 sections, 13 equations, 6 figures, 3 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Outdated Fact Detection
  4. Knowledge Graph Embedding
  5. Preliminaries
  6. Definitions
  7. Task Description
  8. Design of DEAN
  9. Fact Attention Module
  10. Contrastive R2N Graph Module
  11. Training Objective
  12. Complexity Analysis
  13. Experiments
  14. Datasets
  15. Baselines
  16. ...and 4 more sections

Figures (6)

  • Figure 1: An example of outdated facts caused by relation changes
  • Figure 2: The overall framework of DEAN.
  • Figure 3: The computation of attention mechanism.
  • Figure 4: Sensitivities of DEAN w.r.t. K in different datasets.
  • Figure 5: Sensitivities of DEAN w.r.t. $\lambda$
  • ...and 1 more figures