Food Info Harmonized Food Composition Dataset (v1): four national databases in a common schema

A harmonized food-composition dataset that merges four national reference databases into a single common schema, with per-100 g nutrient values and source attribution. Published by Food Info (DCPNET LTD). Most published food-composition tables are national and siloed, each with its own schema, nutrient vocabulary, units, and file format. This dataset's contribution is the harmonization: four official databases brought into one consistent schema and nutrient vocabulary, per 100 g, so a researcher can query across countries without re-deriving the merge themselves. This first release (v1) covers the four databases whose licences unambiguously permit third-party redistribution. Further national databases (e.g. Denmark, Canada, Finland) are planned for later versions under the same concept DOI. SOURCES USDA FoodData Central (SR Legacy, Foundation, FNDDS), United States — 13,658 foods — Public domain (U.S. Government work) CoFID (McCance & Widdowson's Composition of Foods Integrated Dataset), United Kingdom — 2,886 foods — Open Government Licence v3.0 Ciqual (ANSES French food composition table), France — 3,442 foods — Licence Ouverte 2.0 AUSNUT (FSANZ), Australia — 3,735 foods — CC-BY 3.0 AU 23,721 foods in total (reference foods carrying at least 8 reported nutrients), and 1,343,962 nutrient values across the four databases. WHAT'S IN IT Two linked CSV files (see DATA_DICTIONARY.md): foods.csv — one row per food: id, source database, country, name language, name, completeness. food_nutrients.csv — long format, one row per (food, nutrient): id, nutrient, unit, amount per 100 g. METHOD Each national database is ingested with its native identifiers, units, and nutrient definitions preserved. Nutrients are mapped to a common vocabulary and units are normalized so a value means the same thing across sources. All values are expressed per 100 g of edible portion. Foods below the 8-nutrient completeness threshold are excluded so the release is usable rather than sparse. No nutrient values are altered from their source; only the schema, nutrient names, and units are harmonized. LICENCE The harmonized compilation (this dataset's structure, nutrient mapping, and selection) is released under Creative Commons Attribution 4.0 (CC-BY 4.0). Each underlying value remains governed by its source database's own licence (see Sources). Users must retain attribution to the original national databases, not only to Food Info.