It's possible to envision a web/network/tree, or more formally a graph () in which there are three types of nodes: source, ingredient, dish. A source is usually an animal, plant, fungi etc. species. eg. Cattle (species name is "Bos taurus") An ingredient is a component etc. taken from a source. eg. Sirloin steak, a cut of beef from cattle A dish is something that can be consumed directly. (Eg. Fresh milk, caviar, a cake, bread) Many "ingredients" can also be a "dish", under this definition.

The edges (represented by "arrows") would then represent the "is used to make" relation. So Cattle -> Sirloin steak Chicken -> Egg -> Cake (oversimplified.. as there are many kinds of cakes.. but you get the drift..)

By extension, the edges can even be labelled with certain processing, cooking etc. methods. Eg. Pig -> Pork -(salt-cured)-> Bacon

Are there such databases / datasets in the public domain / on the internet?

  • not that I am aware of - that would be one hell of a relational database, given the diversity of dishes and food sources around the world.
    – bob1
    Commented May 29, 2023 at 2:46

1 Answer 1


After a quick search, it seems like Sony AI and Korea University recently started developing an AI being able to predict the mixes between several flavors, FlavorGraph. I read the article and it looks quite powerful, even though it does not entirely fulfill your conditions and is more about "food pairing" (by chemical and molecular associations) than linking dishes to ingredients and/or sources. However, it might be helpful to find some interesting associations between flavors, thus maybe invent dishes, who knows ?

I hope it helps, I will post down below some links related to this subject, so everyone can make an opinion and maybe correct my answer. :) Lovely day !

Edit (2023/05/30, 03:51PM) : I also found a GitHub page providing a FlavorGraph-like code implementation (https://github.com/lamypark/FlavorGraph)

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.