Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Disentangled Multi-span Evolutionary Network against Temporal Knowledge Graph Reasoning

Hao Dong, Ziyue Qiao, Zhiyuan Ning, Qi Hao, Yi Du, Pengyang Wang, Yuanchun Zhou

TL;DR

The paper tackles temporal knowledge graph extrapolation by introducing DiMNet, a disentangled multi-span evolutionary network. The encoder captures cross-time interactions among subgraphs through a multi-span evolution backbone, while a cross-time disentanglement module separates active and stable features to better model semantic changes. A query-aware decoder with inference via sampling a virtual subgraph further mitigates future-topology uncertainty. Empirical results on four benchmarks show state-of-the-art performance and robust improvements over baselines, highlighting the importance of internal subgraph interactions and dynamic feature disentanglement for temporal reasoning.

Abstract

Temporal Knowledge Graphs (TKGs), as an extension of static Knowledge Graphs (KGs), incorporate the temporal feature to express the transience of knowledge by describing when facts occur. TKG extrapolation aims to infer possible future facts based on known history, which has garnered significant attention in recent years. Some existing methods treat TKG as a sequence of independent subgraphs to model temporal evolution patterns, demonstrating impressive reasoning performance. However, they still have limitations: 1) In modeling subgraph semantic evolution, they usually neglect the internal structural interactions between subgraphs, which are actually crucial for encoding TKGs. 2) They overlook the potential smooth features that do not lead to semantic changes, which should be distinguished from the semantic evolution process. Therefore, we propose a novel Disentangled Multi-span Evolutionary Network (DiMNet) for TKG reasoning. Specifically, we design a multi-span evolution strategy that captures local neighbor features while perceiving historical neighbor semantic information, thus enabling internal interactions between subgraphs during the evolution process. To maximize the capture of semantic change patterns, we design a disentangle component that adaptively separates nodes' active and stable features, used to dynamically control the influence of historical semantics on future evolution. Extensive experiments conducted on four real-world TKG datasets show that DiMNet demonstrates substantial performance in TKG reasoning, and outperforms the state-of-the-art up to 22.7% in MRR.

Disentangled Multi-span Evolutionary Network against Temporal Knowledge Graph Reasoning

TL;DR

The paper tackles temporal knowledge graph extrapolation by introducing DiMNet, a disentangled multi-span evolutionary network. The encoder captures cross-time interactions among subgraphs through a multi-span evolution backbone, while a cross-time disentanglement module separates active and stable features to better model semantic changes. A query-aware decoder with inference via sampling a virtual subgraph further mitigates future-topology uncertainty. Empirical results on four benchmarks show state-of-the-art performance and robust improvements over baselines, highlighting the importance of internal subgraph interactions and dynamic feature disentanglement for temporal reasoning.

Abstract

Temporal Knowledge Graphs (TKGs), as an extension of static Knowledge Graphs (KGs), incorporate the temporal feature to express the transience of knowledge by describing when facts occur. TKG extrapolation aims to infer possible future facts based on known history, which has garnered significant attention in recent years. Some existing methods treat TKG as a sequence of independent subgraphs to model temporal evolution patterns, demonstrating impressive reasoning performance. However, they still have limitations: 1) In modeling subgraph semantic evolution, they usually neglect the internal structural interactions between subgraphs, which are actually crucial for encoding TKGs. 2) They overlook the potential smooth features that do not lead to semantic changes, which should be distinguished from the semantic evolution process. Therefore, we propose a novel Disentangled Multi-span Evolutionary Network (DiMNet) for TKG reasoning. Specifically, we design a multi-span evolution strategy that captures local neighbor features while perceiving historical neighbor semantic information, thus enabling internal interactions between subgraphs during the evolution process. To maximize the capture of semantic change patterns, we design a disentangle component that adaptively separates nodes' active and stable features, used to dynamically control the influence of historical semantics on future evolution. Extensive experiments conducted on four real-world TKG datasets show that DiMNet demonstrates substantial performance in TKG reasoning, and outperforms the state-of-the-art up to 22.7% in MRR.

Paper Structure

This paper contains 26 sections, 17 equations, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Definition of TKG
  4. Formulation of Reasoning Task
  5. Methodology
  6. Model Overview
  7. Multi-span Evolution Backbone
  8. Message Aggregation Process
  9. Feature Update after Aggregation
  10. Initialization of the Starting Layer and Timestamp
  11. Cross-time Disentanglement
  12. Decoder and Inference
  13. Score Function
  14. Inference by Sampling Virtual Subgraph
  15. Learning and Optimization
  16. ...and 11 more sections

Figures (5)

  • Figure 1: Illustration of the current central node perceiving historical neighbors. When the central node in the current subgraph aggregates neighbor information, historical neighbor information with the same hop is perceived to enhance the semantic aggregation of neighbors.
  • Figure 2: The Overall Architecture of DiMNet.
  • Figure 3: Performance on Different Sequence Length $m$.
  • Figure 4: Performance on Different GNN Layer Number $\omega$.
  • Figure 5: Performance on Different Virtual Subgraph Sampling Number $k$.