TSOTSAGraph: A Linked Open Food Knowledge Graph

TSOTSAGraph is a multimodal food knowledge graph developed for the field of Food Information Engineering and built on top of the Open Research Knowledge Graph (ORKG) infrastructure. The graph integrates heterogeneous food composition knowledge collected from scientific publications, the USDA Food Composition Database, and FAO reports, with a particular emphasis on African food resources that remain underrepresented in existing food knowledge graphs.

The current release contains 542,665 RDF triples describing 76,956 named individuals through 46 ontology classes and 41 semantic properties. The graph captures semantic relationships between foods, ingredients, nutrients, food composition data, and provenance information, providing a structured and interoperable representation of food knowledge.

TSOTSAGraph currently describes 1,818 food entities, including 350 raw foods (82 fruits and 268 vegetables), 1,468 processed and prepared foods, and 24 beverages. The resource further includes 847 food ingredients and 273 food components. Among the represented foods, 1,624 originate from African food composition resources, while 194 are aligned with USDA food records.

Food entities are represented through an ontology-driven model that captures foods, ingredients, nutrients, food composition data, and associated metadata. The graph can be used for semantic retrieval, food composition analysis, data integration, and other knowledge-based applications in the food domain.

The dataset is distributed in RDF/Turtle format and follows the ORKG knowledge representation model, facilitating future extensions and the integration of additional food composition resources.