The three layers of a knowledge graph and what it means for authoring, storage, governance and data quality of the knowledge graph
0 likes474 views
The document discusses the three layers of a knowledge graph: the ontology, taxonomies, and knowledge. It explains that the ontology contains the core business concepts, taxonomies contain controlled vocabularies, and knowledge contains the network of products. It then covers authorship, storage, governance, and data quality considerations for each layer. Authorship can be centralized or decentralized, storage options include triple stores and property graphs, governance involves peer and expert reviews, and data quality is ensured through validation and enabling use cases across layers.
The three layers of a knowledge graph and what it means for authoring, storage, governance and data quality of the knowledge graph
- 1. © Inter IKEA Systems B.V. 2023 © Inter IKEA Systems B.V. 2023 1 The Three Layers of a Knowledge Graph Katariina Kari Inter IKEA Systems B.V.
- 2. © Inter IKEA Systems B.V. 2023 © Inter IKEA Systems B.V. 2023 2 2 ThreeLayersof a KnowledgeGraph and what it means for authoring, storage, governance and data quality.
- 3. © Inter IKEA Systems B.V. 2022 Internal 3 Agenda Overview of the Knowledge Graph A Game of 20 Questions Authorship Storage Governance Data Quality
- 5. © Inter IKEA Systems B.V. 2023 Internal 5 A new infrastructure combining business concepts with actual data in a connected network-like form. This allows for richer insights around products interior design! What the knowledge graph is... and why I like them!
- 6. © Inter IKEA Systems B.V. 2023 Internal 6 Business concepts central to IKEA (schema) The controlled vocabulary (instances) All products in network structure (also instances)
- 7. © Inter IKEA Systems B.V. 2023 Internal 7 Business concepts central to IKEA = ontology The controlled vocabulary = taxonomies All products in network structure = knowledge
- 8. © Inter IKEA Systems B.V. 2022 Internal 8 A Gameof 20 Questions Try it at home: Create a Google Form for a game of 20 questions to answer what furniture a player is thinking of.
- 12. © Inter IKEA Systems B.V. 2023 Internal 12 Centralised and owned by the ontology team Decentralised and owned by expert teams. Automatically generated from a data source. Ownership is at the source.
- 14. © Inter IKEA Systems B.V. 2023 Internal 14 RDF file, TTL syntax in a Git repo RDF file, TTL syntax in a Git repo Triple Store for Browsing and Faceted Search Labelled Property Graph for path queries.
- 16. © Inter IKEA Systems B.V. 2023 Internal 16 Review by ontologist peers and negotiation with domain experts Final say by domain expert teams. No governance, since it is generated.
- 18. © Inter IKEA Systems B.V. 2023 Internal 18 Usability of queries, reasoning capabilities, enabling use cases. Reusability for use cases across. SHACL validation.
- 19. Semantic and Transparent ETL Pipelines - Katariina Kari, Lead Ontologist, Inter IKEA Systems B.V. Questions? © Inter IKEA Systems B.V. 2022