I would be more concerned about accuracy, bias and curation sooner than I would think about the license in case of a community-built dataset. Take a website such as TasteAtlas, for example. They frequently publish rankings of best/worst foodstufs, but their data on that is garbage (even if you skip the fact that taste is highly subjective). Not only are there lots of incorrect assumptions made (e.g. a dish always belongs to only one national cuisine), but also the popularity data is not curated at all, so they end up with situations such as the first place in my country on the "100 Best Food Destinations" list was a tiny village with a truck stop.