Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Temporal Knowledge Graph Hyperedge Forecasting: Exploring Entity-to-Category Link Prediction

Edward Markai, Sina Molavipour

TL;DR

This paper tackles the interpretability gap in temporal knowledge graph forecasting by extending a rule-based framework, TLogic, to a category-aware variant, C-TLogic. It formalizes category-inclusive rules using the sextuple format $(s,r,o,t,c^s,c^o)$ and introduces a data-driven approach to assign entity categories, enabling both entity- and category-level forecasts. Rules are mined via temporal random walks, grounded to data, and scored with $f(R,k)=\alpha\,conf(R)+(1-\alpha)\exp(-\lambda (t_q-\tau))$, then aggregated with Noisy-OR or Max+ to form final predictions. Empirical results on ICEWS18 and FinDKG show that while TLogic generally yields stronger entity-level predictions, the category-aware approach reveals trade-offs in granularity and generalization, with results highly dependent on dataset structure and categorization quality.

Abstract

Temporal Knowledge Graphs have emerged as a powerful way of not only modeling static relationships between entities but also the dynamics of how relations evolve over time. As these informational structures can be used to store information from a real-world setting, such as a news flow, predicting future graph components to a certain extent equates predicting real-world events. Most of the research in this field focuses on embedding-based methods, often leveraging convolutional neural net architectures. These solutions act as black boxes, limiting insight. In this paper, we explore an extension to an established rule-based framework, TLogic, that yields a high accuracy in combination with explainable predictions. This offers transparency and allows the end-user to critically evaluate the rules applied at the end of the prediction stage. The new rule format incorporates entity category as a key component with the purpose of limiting rule application only to relevant entities. When categories are unknown for building the graph, we propose a data-driven method to generate them with an LLM-based approach. Additionally, we investigate the choice of aggregation method for scores of retrieved entities when performing category prediction.

Temporal Knowledge Graph Hyperedge Forecasting: Exploring Entity-to-Category Link Prediction

TL;DR

This paper tackles the interpretability gap in temporal knowledge graph forecasting by extending a rule-based framework, TLogic, to a category-aware variant, C-TLogic. It formalizes category-inclusive rules using the sextuple format and introduces a data-driven approach to assign entity categories, enabling both entity- and category-level forecasts. Rules are mined via temporal random walks, grounded to data, and scored with , then aggregated with Noisy-OR or Max+ to form final predictions. Empirical results on ICEWS18 and FinDKG show that while TLogic generally yields stronger entity-level predictions, the category-aware approach reveals trade-offs in granularity and generalization, with results highly dependent on dataset structure and categorization quality.

Abstract

Temporal Knowledge Graphs have emerged as a powerful way of not only modeling static relationships between entities but also the dynamics of how relations evolve over time. As these informational structures can be used to store information from a real-world setting, such as a news flow, predicting future graph components to a certain extent equates predicting real-world events. Most of the research in this field focuses on embedding-based methods, often leveraging convolutional neural net architectures. These solutions act as black boxes, limiting insight. In this paper, we explore an extension to an established rule-based framework, TLogic, that yields a high accuracy in combination with explainable predictions. This offers transparency and allows the end-user to critically evaluate the rules applied at the end of the prediction stage. The new rule format incorporates entity category as a key component with the purpose of limiting rule application only to relevant entities. When categories are unknown for building the graph, we propose a data-driven method to generate them with an LLM-based approach. Additionally, we investigate the choice of aggregation method for scores of retrieved entities when performing category prediction.

Paper Structure

This paper contains 16 sections, 16 equations, 2 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Problem
  3. Method
  4. Assigning categories
  5. Temporal Logic Rules
  6. Preliminaries
  7. Rule Learning
  8. Rule application
  9. Candidate aggregation
  10. Implementation
  11. Learning
  12. Retriever
  13. Evaluation
  14. Results and Analysis
  15. Discussion
  16. ...and 1 more sections

Figures (2)

  • Figure 1: Subset of top 10 entities by degree at timestamps 2018-11-04 (above), and 2018-11-06 (below) in the ICEWS18_12 TKG. The color of the nodes indicate different entity categories.
  • Figure 2: Frequency of rules per body support level for rules trained on the datasets ICEWS18_12 (top) and FinDKG_Comp (bottom), when training TLogic (left) and C-TLogic (right).