A new research paper introduces FedV-KGQA, a system designed for federated knowledge graph question answering over vertically partitioned graphs. This approach addresses scenarios where data is distributed across organizations that share entity identifiers but have disjoint relation types, preventing any single entity from accessing the complete graph. The system enriches local graphs, trains embeddings within silos, and then concatenates silo-specific entity views on a server for question anchoring and candidate ranking, ensuring raw triples and relation embeddings never leave their respective silos. Experiments show that federated fusion significantly recovers centralized accuracy, with anchoring and enrichment proving more critical than the choice of embedding model. AI
IMPACT This research could enable more effective question answering over distributed knowledge graphs, improving data access and analysis for organizations with siloed information.
RANK_REASON The cluster contains a research paper published on arXiv detailing a new system and its empirical findings.
- alphaXiv
- arXiv
- CatalyzeX
- DagsHub
- FedV-KGQA
- Gotit.pub
- Hugging Face
- Knowledge Graph Question Answering
- Md Saikat Islam Khan Bappy
- ScienceCast
- Influence Flower
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →