Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

HIP Network: Historical Information Passing Network for Extrapolation Reasoning on Temporal Knowledge Graph

Yongquan He, Peng Zhang, Luchen Liu, Qi Liang, Wenyuan Zhang, Chuang Zhang

TL;DR

Extrapolation reasoning on temporal knowledge graphs is challenging because future graphs are unseen during training. The HIP network addresses this by passing historical information from temporal, structural, and repetitive perspectives, updating relation representations with a disentangled CompGCN, modeling pairwise temporal evolution with temporal self-attention and GRUs, and using three scoring functions to predict future events step by step. Empirical results on five benchmarks show state-of-the-art performance, with notable gains in Hit@1 and MRR, demonstrating the value of multi-perspective history-aware reasoning. This approach enhances forecasting of future events in dynamic knowledge graphs and provides a scalable framework for leveraging historical patterns and relation dynamics in extrapolation tasks.

Abstract

In recent years, temporal knowledge graph (TKG) reasoning has received significant attention. Most existing methods assume that all timestamps and corresponding graphs are available during training, which makes it difficult to predict future events. To address this issue, recent works learn to infer future events based on historical information. However, these methods do not comprehensively consider the latent patterns behind temporal changes, to pass historical information selectively, update representations appropriately and predict events accurately. In this paper, we propose the Historical Information Passing (HIP) network to predict future events. HIP network passes information from temporal, structural and repetitive perspectives, which are used to model the temporal evolution of events, the interactions of events at the same time step, and the known events respectively. In particular, our method considers the updating of relation representations and adopts three scoring functions corresponding to the above dimensions. Experimental results on five benchmark datasets show the superiority of HIP network, and the significant improvements on Hits@1 prove that our method can more accurately predict what is going to happen.

HIP Network: Historical Information Passing Network for Extrapolation Reasoning on Temporal Knowledge Graph

TL;DR

Extrapolation reasoning on temporal knowledge graphs is challenging because future graphs are unseen during training. The HIP network addresses this by passing historical information from temporal, structural, and repetitive perspectives, updating relation representations with a disentangled CompGCN, modeling pairwise temporal evolution with temporal self-attention and GRUs, and using three scoring functions to predict future events step by step. Empirical results on five benchmarks show state-of-the-art performance, with notable gains in Hit@1 and MRR, demonstrating the value of multi-perspective history-aware reasoning. This approach enhances forecasting of future events in dynamic knowledge graphs and provides a scalable framework for leveraging historical patterns and relation dynamics in extrapolation tasks.

Abstract

In recent years, temporal knowledge graph (TKG) reasoning has received significant attention. Most existing methods assume that all timestamps and corresponding graphs are available during training, which makes it difficult to predict future events. To address this issue, recent works learn to infer future events based on historical information. However, these methods do not comprehensively consider the latent patterns behind temporal changes, to pass historical information selectively, update representations appropriately and predict events accurately. In this paper, we propose the Historical Information Passing (HIP) network to predict future events. HIP network passes information from temporal, structural and repetitive perspectives, which are used to model the temporal evolution of events, the interactions of events at the same time step, and the known events respectively. In particular, our method considers the updating of relation representations and adopts three scoring functions corresponding to the above dimensions. Experimental results on five benchmark datasets show the superiority of HIP network, and the significant improvements on Hits@1 prove that our method can more accurately predict what is going to happen.
Paper Structure (23 sections, 14 equations, 3 figures, 3 tables, 1 algorithm)

This paper contains 23 sections, 14 equations, 3 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Static KG embeddings.
  4. Temporal KG reasoning.
  5. Method
  6. Task Definition and Model Architecture
  7. Notations and task definition.
  8. Model architecture.
  9. Structural Information Passing Module
  10. Temporal Information Passing Module
  11. History Forward Module
  12. Training Objective
  13. Experiments
  14. Experimental Setup
  15. Datasets.
  16. ...and 8 more sections

Figures (3)

  • Figure 1: Temporal knowledge subgraphs about the evolution of political events in Syria and the Trump government's responses.
  • Figure 2: The framework of our HIP network. HIP network contains three components. The structural information passing module to capture neighborhood interactions. The temporal information passing module to model evolution patterns of events. And the HIP module to calculate the plausibility of events from the repetitive, structural and temporal perspectives.
  • Figure 3: Parameter sensitivity on HIP network.