非欧几里得空间中的神经符号逻辑查询回答

The proposed model, HYQNET, addresses the challenge of answering complex first-order logic (FOL) queries on knowledge graphs, which is crucial for reasoning. While symbolic methods provide interpretability, they struggle with incomplete graphs. Neural approaches, on the other hand, generalize better but lack transparency. Neural-symbolic models aim to combine these strengths but often fail to capture the hierarchical structure of logical queries, limiting their effectiveness. HYQNET leverages hyperbolic space for logic query reasoning by decomposing FOL queries into relation projections and logical operations over fuzzy sets, thus enhancing interpretability. To tackle missing links, it employs a hyperbolic GNN-based approach for knowledge graph completion in hyperbolic space.