TSOTSAGraph: A Linked Open Food Knowledge Graph

TSOTSAGraph is a linked food knowledge graph developed for Food Information Engineering on top of the Open Research Knowledge Graph (ORKG). It integrates food knowledge curated from scientific publications and food composition resources, with particular emphasis on African foods that remain underrepresented in existing food knowledge resources.

Version 2.1.0 contains 994,965 RDF triples, 140,319 RDF subjects, 38 declared classes, 26 object properties, 14 datatype properties, and 59,834 named individuals. It describes 1,834 food resources, including 24 beverages, 82 fruits, 293 vegetables, and 1,435 processed or prepared foods. The graph also contains 2,236 ingredient resources representing 846 distinct ingredient labels and 273 food components.

This release distinguishes the immutable ORKG source layer from the cleaned and enriched graph. It includes normalized quantity values, explicit provenance links, OWL declarations, domain and range axioms, ORKG template constraints represented with SHACL, and local quality constraints.

Version 2.1.0 introduces reviewed semantic alignments with FoodOn, FoodKG, ChEBI, the NCI Thesaurus, and USDA FoodData Central. The seven mapping layers contribute 22,886 triples and support interoperability, entity linking, knowledge-graph completion, and grounded nutritional question answering.

The release archive contains the exact input snapshot, the source-only ORKG graph, the final enriched graph, all mapping layers, detailed statistics, a machine-readable manifest, and SHA-256 checksums. The standalone Turtle file is provided for direct loading into RDF systems such as Apache Jena Fuseki and Protégé.

The source code and updated documentation are available at: TSOTSAGraph download