Formalizing Property Constraints in Wikidata. Ferranti, N., Polleres, A., Souza, J. F. D., & Ahmetaj, S. In Proceedings of the 3rd Wikidata Workshop (co-located with ISWC2022), October, 2022.
Formalizing Property Constraints in Wikidata [pdf]Paper  abstract   bibtex   
Constraints play an important role to ensure data integrity. While the Shapes Constraint Language (SHACL) provides a W3C recommendation for validating RDF Knowledge Graphs (KG) against such constraints, real-world KG have adopted their own constraint formalisms. Wikidata (WD), one of the largest collaboratively Open Data Knowledge Graphs available on the Web, represents property constraints through its own RDF data model, within its own authoritative namespaces, which might be an indication that the nature of WD property constraints is different from other Knowledge Graphs. In this paper we investigate the semantics of WD constraints, and unambiguously formalize all current constraints using SPARQL to retrieve violations; we also discuss the expressiveness of WD constraint language compared with SHACL core and discuss the evolution of constraint violations. We found that, while all current WD property constraint types can be expressed using SPARQL, only 86% (26 out of 30) can be expressed using SHACL core: the rest face issues related to using separator properties and arithmetic expressions.

Downloads: 0