is a Knowledge Graph.
FoodKG is a food knowledge graph built at Rensselaer Polytechnic Institute as part of the HEALS project with IBM Research. It links about one million recipes from the Recipe1M dataset, their ingredients, USDA nutrient data and the FoodOn food ontology, with provenance for each statement, to support food recommendation and question answering. The project releases its build scripts, mapping files and supporting ontologies rather than full RDF dumps; the full graph (about 67 million triples) is generated locally or queried through a public SPARQL endpoint, which returned HTTP 502 when checked on 2026-10-04.
Unknown
No external IDs available
| ID | Name | URL | Category | Format | Description |
|---|---|---|---|---|---|
| foodkg.site | FoodKG Website | foodkg.github.io | Documentation | http | FoodKG project website, describing th... |
| foodkg.build-scripts | FoodKG Construction Scripts | src | Process | python | Python scripts that build FoodKG from... |
| foodkg.sparql ⚠ | FoodKG SPARQL Endpoint | sparql | Programming Interface | http | Public Blazegraph SPARQL endpoint for... |
| foodkg.usda-sample | FoodKG USDA Mappings Sample | open?id=1hkitCcxnM_7R6OYuvC5zakWojlN2Xuog | Graph | rdfxml | Sample of FoodKG containing the USDA ... |
| foodkg.sdd-dictionary | FoodKG USDA Semantic Data Dictionary | usdaDM.csv (1.0 KB) | Mapping | csv | Semantic Data Dictionary mapping file... |
| foodkg.whattomake | WhatToMake Ontology | food (6.4 KB) | Ontology | rdfxml | Food component of the WhatToMake onto... |
| foodkg.whattomake-foodon | WhatToMake FoodOn Module | foodon (14.9 KB) | Ontology | rdfxml | FoodOn component of the WhatToMake on... |
| foodkg.whattomake-individuals | WhatToMake Individuals | ingredient (14.0 KB) | Graph | rdfxml | Ingredient component of the WhatToMak... |
| foodkg.dgo | Dietary Guideline Ontology | dgo.owl (3.6 KB) | Ontology | ttl | Dietary Guideline Ontology (heals-gui... |
FoodKG is a food knowledge graph developed by the Tetherless World Constellation at Rensselaer Polytechnic Institute, with IBM Research, as part of the Health Empowerment by Analytics, Learning and Semantics (HEALS) project. It was presented at ISWC 2019 as a resource for consumer-oriented food recommendation.
FoodKG combines three kinds of data, keeping provenance for each statement:
The project publishes construction scripts rather than full RDF dumps. Users obtain the Recipe1M files themselves and run the scripts to produce usda-links.trig, foodon-links.trig and foodkg-core.trig (about 67 million triples in total), which can be loaded into Blazegraph. A sample of the USDA mappings is available as RDF/XML, and the site documents a public SPARQL endpoint, which returned HTTP 502 when checked on 2026-10-04.
The site also hosts the WhatToMake ontology, the Dietary Guideline Ontology, and related work on question answering and ingredient substitution over FoodKG. The repository was last updated in November 2024.
Created: October 04, 2026 | Last modified: October 05, 2026