[en] Purpose: This paper addresses the lack of a general declarative approach for generating Wikibase content from heterogeneous data sources. Existing declarative mapping approaches mainly target Resource Description Framework (RDF), while Wikibase-oriented solutions are often limited in scope or procedural in nature. We therefore propose the Wikibase Mapping Language (WBML), a declarative mapping language tailored to the Wikibase statement model. Methodology: WBML was designed by reusing selected RML concepts for logical sources and iterations, while introducing Wikibase-specific constructs. We implemented a Python prototype that translates WBML into intermediate RDF Mapping Language (RML) and RDF representations, executes them with Morph-KGC, and ingests the results into Wikibase. Findings: The study shows that declaratively generating Wikibase statements from heterogeneous sources is feasible. We also reported on a demonstration using complex JSON and a simpler, real-world use case for our institution, using CSV files. WBML supports the generation of entities, fingerprints, statements, qualifiers, references, and ranks. Value: This work provides a basis for declarative construction of Wikibase knowledge graphs from heterogeneous sources. It demonstrates how established ideas from the RDF mapping community can be adapted to other types of knowledge graphs. Future work includes supporting newer RML features not yet supported by reused components and managing updates.
Disciplines :
Computer science
Author, co-author :
Liu, Elena ; Université de Liège - ULiège > Montefiore Institute of Electrical Engineering and Computer Science
Duchateau, Jakub ; Université de Liège - ULiège > Montefiore Institute of Electrical Engineering and Computer Science
Debruyne, Christophe ; Université de Liège - ULiège > Montefiore Institute of Electrical Engineering and Computer Science