Learning Collaborative Agents with Rule Guidance for Knowledge Graph Reasoning

TL;DR

This paper introduces RuleGuider, a method that combines symbolic rule guidance with walk-based reinforcement learning to improve knowledge graph reasoning, enhancing performance while maintaining interpretability.

Contribution

It proposes a novel framework that integrates symbolic rules as reward signals to guide walk-based KG reasoning models, addressing sparse reward issues.

Findings

RuleGuider outperforms baseline walk-based models on benchmark datasets.

The approach maintains interpretability of reasoning paths.

Enhanced reasoning accuracy demonstrated across multiple datasets.

Abstract

Walk-based models have shown their advantages in knowledge graph (KG) reasoning by achieving decent performance while providing interpretable decisions. However, the sparse reward signals offered by the KG during traversal are often insufficient to guide a sophisticated walk-based reinforcement learning (RL) model. An alternate approach is to use traditional symbolic methods (e.g., rule induction), which achieve good performance but can be hard to generalize due to the limitation of symbolic representation. In this paper, we propose RuleGuider, which leverages high-quality rules generated by symbolic-based methods to provide reward supervision for walk-based agents. Experiments on benchmark datasets show that RuleGuider improves the performance of walk-based models without losing interpretability.