Abstract:The incompleteness of knowledge graphs (KGs) negatively impacts the performance of KGs in downstream applications (e.g., recommendation systems and information retrieval). This phenomenon has brought an increasing rise in research related to knowledge graph reasoning. Recently, emerged reinforcement learning (RL)-based multi-hop reasoning methods can infer missing information through multi-hop reasoning according to the existing information in KGs, which has better reasoning performance and interpretability. However, these methods always use relation-entity pairs that have been pre-cropped as the action space of agents for path reasoning, which leads to two problems: 1) insufficient learning and reasoning ability of reasoning models and 2) the hard convergence of the training process of agents. To address these problems, we propose a Hierarchical Sequence Multi Hop (HSMH) reasoning framework, which consists of the interactive search reasoning model, local-global knowledge fusion mechanism, and action optimization mechanism. We use interactive search reasoning models to select relations and entities independently, thus fully mining the semantic information of relations and entities and improving the learning and reasoning ability of reasoning models. In the HSMH framework, we design the local-global knowledge fusion and action optimization mechanisms for path reasoning, which can enhance agents' state information and action space. Specifically, the local-global knowledge fusion mechanism is designed to acquire the local knowledge of entities and neighboring relations and the global knowledge about KG structure. This local-global knowledge can improve the learning ability of reasoning models. In addition, the action optimization mechanism can combine the filtered action space and the additional action space for efficient path reasoning for agents. Experimental results on five benchmark datasets show that our proposed HSMH framework comprehensively outperforms the state-of-the-art multi-hop reasoning model.

Fusion of Natural Language and Knowledge Graph for Multi-hop Reasoning

MMKGR: Multi-hop Multi-modal Knowledge Graph Reasoning

GMH: A General Multi-hop Reasoning Model for KG Completion

Neural Axiom Network for Knowledge Graph Reasoning

Multi-hop Knowledge Reasoning with Deep Reinforcement Learning

Unrestricted multi-hop reasoning network for interpretable question answering over knowledge graph

Adapting Meta Knowledge Graph Information for Multi-Hop Reasoning over Few-Shot Relations

ExKGR: Explainable Multi-hop Reasoning for Evolving Knowledge Graph

HSMH: A Hierarchical Sequence Multi-Hop Reasoning Model With Reinforcement Learning

Incorporating multi-perspective information into reinforcement learning to address multi-hop knowledge graph question answering

Attention-based Exploitation and Exploration Strategy for Multi-Hop Knowledge Graph Reasoning

MKGN: A Multi-Dimensional Knowledge Enhanced Graph Network for Multi-Hop Question and Answering

HackRL: Reinforcement learning with hierarchical attention for cross-graph knowledge fusion and collaborative reasoning

Learning Multi-Granularity and Adaptive Representation for Knowledge Graph Reasoning

Multi-Hop Knowledge Graph Reasoning in Few-Shot Scenarios

Step by Step: A Hierarchical Framework for Multi-Hop Knowledge Graph Reasoning with Reinforcement Learning

HiAM: A Hierarchical Attention based Model for knowledge graph multi-hop reasoning

UniKGQA: Unified Retrieval and Reasoning for Solving Multi-hop Question Answering Over Knowledge Graph

CogKR: Cognitive Graph for Multi-hop Knowledge Reasoning

Relgraph: A Multi-Relational Graph Neural Network Framework for Knowledge Graph Reasoning Based on Relation Graph

MCMH: Learning Multi-Chain Multi-Hop Rules for Knowledge Graph Reasoning