Learning Collaborative Agents with Rule Guidance for Knowledge Graph Reasoning
This work addresses a bottleneck in knowledge graph reasoning for AI applications, offering an incremental improvement by combining symbolic and reinforcement learning methods.
The paper tackles the problem of sparse reward signals in walk-based knowledge graph reasoning by proposing RuleGuider, which uses symbolic rules to provide reward supervision, resulting in improved performance on benchmark datasets while maintaining interpretability.
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.