跳转至

FedV-KGQA：基于垂直划分知识图谱的多跳问答框架

文章背景与核心概要

现实世界中的知识图谱问答（KGQA）常常受到组织壁垒和数据主权约束的阻碍。当所需的事实分散在不同地方时，往往难以回答复杂的多跳问题。为此，本文介绍了 FedV-KGQA 这一新颖的框架，旨在解决跨垂直划分知识图谱（即各组织共享实体但拥有不相交的关系集）的多跳推理难题。

该框架通过结合本地图谱丰富技术（Local Graph Enrichment）与知识图谱嵌入，在不暴露原始三元组或关系参数的前提下，建立了强有力的结构化数据边界。它利用主题实体锚定机制，在本地将问题定位到合适的图邻域中，从而消除了运行时的跨孤岛通信需求，同时达到了接近集中式方法的性能水平。

FedV-KGQA: Multi-Hop Question Answering over Vertically Partitioned Knowledge Graphs

Real-world knowledge graph question answering (KGQA) is frequently hindered by organizational silos and data sovereignty constraints, making it difficult to answer multi-hop questions when required facts are distributed. FedV-KGQA is a novel framework designed to perform multi-hop reasoning over vertically partitioned knowledge graphs—where organizations share entities but own disjoint relation sets.

By combining local graph enrichment and knowledge graph embeddings without exposing raw triples or relation parameters, FedV-KGQA establishes a strong structural data boundary. It utilizes a topic entity anchoring mechanism to ground questions in the appropriate graph neighborhood locally, eliminating the need for runtime inter-silo communication while achieving performance close to centralized approaches.

元数据与出版详情

Metadata & Publication Details

  • arXiv 标识符： arXiv:2608.24846 [cs.AI]

  • 作者： Md Saikat Islam Khan Bappy, Oshani Seneviratne

    • Authors: Md Saikat Islam Khan Bappy, Oshani Seneviratne

  • 提交时间： 2026年8月25日

    • Submitted: August 25, 2026

  • 录用情况： 已被 ISWC 2026（研究赛道）录用，即将发表于第25届国际语义网会议论文集

    • Accepted at: ISWC 2026 (Research Track), to appear in the Proceedings of the 25th International Semantic Web Conference

摘要

Abstract

由于治理和数据主权限制，用于知识图谱问答的现实世界数据往往分布在不同的组织中。虽然存在集中式系统，但当所需的事实被分割在垂直划分的孤岛中时，它们无法回答多跳问题。

Real-world data for knowledge graph question answering is often distributed across different organizations due to governance and data sovereignty constraints. While centralized systems exist, they cannot answer multi-hop questions when the required facts are split across vertically partitioned silos.

在本文中，我们提出了 FedV-KGQA，这是一个用于知识图谱多跳推理的框架，其中各组织共享实体但拥有不相交的关系集。我们的方法结合了本地图谱丰富技术和知识图谱嵌入，以确保原始三元组和关系参数永不离开各个孤岛，从而在无需集中式图谱访问的情况下建立结构化数据边界。我们进一步引入了一种主题实体锚定机制，在没有任何运行时跨孤岛通信的情况下，将问题锚定在正确的图邻域中。我们对三个基准测试中的 12 种模型配置进行了评估，结果表明 FedV-KGQA 表现强劲，性能接近集中式方法，能够泛化到 3 跳推理，并且对嵌入扰动具有鲁棒性。

In this paper, we propose FedV-KGQA, a framework for multi-hop reasoning over knowledge graphs in which organizations share entities but own disjoint sets of relations. Our approach combines local graph enrichment and knowledge graph embeddings to ensure raw triples and relation parameters never leave each silo, establishing a structural data boundary without requiring centralized graph access. We further introduce a topic entity anchoring mechanism that grounds questions in the correct graph neighborhood without any runtime inter-silo communication. We evaluate 12 model configurations across three benchmarks and show that FedV-KGQA performs strongly, remains close to centralized performance, generalizes to 3-hop reasoning, and is robust to embedding perturbations.

访问链接与资源