Subscribe
Sign in
Dish Dragon AI — Data science based ingredient pairing
The results of a survey of 150,000+ online recipes. Find ingredients that combine well or poorly or follow the 3d graph of interlinked ingredients. Also: find recipes based on the Dish Dragon's own scoring system, rather than questionable online review scores.
25
Replies
Tom Frazier
Tried this. Love it.
Share
Nice one! Hacked on something similar mining AllRecipes in the past. Would love to know if you can categorize/filter on dish type.
@jordanesbin Thanks! Technically, I can. I have that in the meta data and could add it. I just have to figure out a nice way to do so UI wise.
I really love the idea of this!
@matseinarsen, I'm curious if you've personally used it to make any dishes, or figured out what to make from an obscure ingredient?
Have you also thought about sorting the pairings into flavour profiles of the combination of ingredients, or of the linked recipes themselves? I'm just wondering if say, you had an ingredient like Truffles, it might be quite interesting to see if there was a clever recipe for a sweet dessert or something, which involved truffles.
@calum Hey! Thanks for the enthusiasm :) Yes - I've found quite a few things that I've liked. The combination of maple syrup and lime juice was really surprising to me, but now I use that in everything from Indian curries to cocktails. Another thing I recently enjoyed was having red onion on blue cheese. I also have it high on my list to splash some honey into my adobo sauce next time I make that. I use it quite a lot myself, actually, but i guess often it's more a reminder of things I already knew, like adding parsley or a chunk of butter to a dish to give it something extra.
Yes on the second :) I already have category meta data for most ingredients and recipes, so that's more an interface thing to add... But it would be interesting to also use a clustering algorithm to identify other kinds of flavour profiles. Definitively pondering that :)
This is a straight forward garage product I've been tinkering on for a while to give my own cooking an edge. I use it to find ingredient pairings that will give what I cook a boost, and I use it to check if something is a bad idea before I add something I'm not sure about.
I also use it to find recipes through a different channel than Google: I want recipes from people who are good at cooking, not SEO.
It's based on 150,000 online recipes and review scores from about 200 different websites, and it's basically just a large database of the average review score for recipes that contain a particular pair or triple of ingredients. As online recipe review scores are quite unreliable (to say the least), I've also put a ton of work into identifying signals in the noise.
