I think you're predicting some interesting future technical problems with Parsnip, which is to build a personalized user-specific knowledge graph that is simultaneously overlaid across generalized common knowledge elements.
One thing at a time, though! This is just an MVP.
Ideally one would be able to learn language with the same kind of etymological study, however I put it to you that is FAR more complex and boring and doesn't have more fact / science based rules than cooking does.
Also, cooking is a very personal / familial / cultural experience. Recipes can be passed down a single family, whereas language is much more broad.
I guess what I'm trying to say is that the knowledge graph for cooking is smaller and more personal, and is better served as a personalized note taking system as opposed to a gamified interface to a public database.
For example, if I would prefer recipes with eggs and potatoes as opposed to recipes with spinach and mushrooms, I should be able to cater my knowledge graph to learning more about recipes that use those ingredients (ideally with some balance in nutrients).