Computerized Cooking (1985)
diyhpl.us
diyhpl.us
I think it's a non-trivial problem. Even with a relatively limited set of go-to dishes and ingredients, it seems to me like the data entry alone would be a herculean effort: making sure everything is normalized, operations like chop vs. mince vs. dice are standardized, etc. Currently, I'm relying on my fragile meat brain and a folder full of text files, skimming a directory listing of the folder for inspiration, and using individual recipes to build shopping lists.
I totally agree with the premise of the original article, that you can reduce recipes to trees and describe them with a formal language, but once I take off my computer scientist hat and look at this behavior critically, it seems horrifically asinine.
Anyone have a better solution? I know there are a couple of open ones out there, and I'm sure some of you nerds have homegrown something way more sophisticated. I've also got a hunch that this would be a great problem to throw language models at...
That said I've been leaning towards a jazzier and improvised kind of cookery. But now I know where to look if I want to logically encode it.
For example, if anyone cares, I've recently added the pasta-beans-cheese meta recipe to my repertoire after seeing it played out a few times in different guises. I think this where the real power lies, generalizing, right?
I think the tree idea is good, as it shows what steps can be done in parallel. It would also be helpful to show how long each subtree is expected to take with working time and waiting time.