FABRIC Knowledge Graph Integration Strategy

The team developed a strategic framework to address two core challenges in knowledge graph integration: how to prioritize resources and entities for mapping and validation, and how to overcome divergent identity schemes to ensure referential consistency. Their approach identifies criteria for prioritization including entity frequency in graphs and alignment with established standards. The framework proposes monitoring query logs for prioritization, mapping broader concepts first for guidance, and using well-established resources like WikiData as pivot points. The team outlined exploration paths including soft standardization via conceptual feature spaces, calculating neighborhood similarity of concepts across resources, and developing tooling to identify quality based on standards compliance and completeness.

The team analyzed the problem space through discussion and identified practical strategies for knowledge graph mapping. They evaluated criteria such as informativity and standard alignment for prioritization decisions. Their approach involved examining manual mapping processes and proposing algorithm-based solutions that leverage conceptual feature spaces rather than rule-based or string-based methods. They focused on qualitative error analysis and explored methods for identifying potential synonyms and measuring concept similarity across different knowledge graph resources.

This work supports Proto-OKN projects and AI-focused research teams working to transition their knowledge graphs to industry applications. The framework addresses the practical challenge of integrating knowledge graphs at scale when manual mapping is required, helping researchers enhance data quality and accelerate the impact of AI research on industry.