Tutorial: Developing and Refining Schemas for Knowledge Graphs

An advantage of modeling data as a graph – as opposed to a relational data model – is that data graph does not require a data schema from the very beginning. Hence graph data is often loaded from external sources with ad-hoc mappings. However, a Knowledge Graph is mostly understood as a data graph enhanced with a schema, a mechanism to distinguish between data and meta-data and sometimes even axioms/rules. Schemata are of great benefit as they describe the structure and semantics of the data in the graph. KG schemata enable

• Different views at various levels of abstraction for a better overview,
• Explicit and therefore exchangeable meaning of the graph data,
• Higher query performance,
• Reasoning by means of axioms/rules.

The tutorial will explain the various schema modeling options for Labeled Property as well as RDF graph data models. We describe the different ways of representing schema information in both models and whether those models are embedded in the data graph or layered above it. Furthermore, we discuss similarities and mutual correspondences as well as advantages and shortcomings of the two graph models. The tutorial will demonstrate tools for the development, maintenance and review of schemas and exemplarily show how to develop and refine KG schemata in both models. In addition, attendees will have the opportunity to work with available KGs and tools to get hands-on experience.