How we did it
The recipes come from RecipeNLG (Bień and others, 2020), 2,231,142 recipes gathered from recipe websites, each with a list of ingredient names that the researchers pulled from the original ingredient lines. We removed recipes with the same name and the same ingredients, which left 2,089,091. This page publishes counts, and the finder shows only each recipe's name and a link to the site it came from.
We tidied the ingredient names by removing words like chopped, fresh and large, turning plurals into singulars and merging common duplicates, so unsalted butter counts as butter and catsup counts as ketchup. Some things a cook would call the same are still counted separately, like chicken and chicken breast, or banana and mashed bananas. Water is left out. Dish names were grouped after removing words like easy, best, homemade and Grandma's. Brand names were matched in the original ingredient text.
Oven temperatures are numbers between 200 and 550 followed by a degree sign or the word degrees in the method. "The closest pairs" compares how often two ingredients appear together with how often they would if they were independent. "How far your kitchen gets you" adds ingredients one at a time, always picking the one that completes the most recipes, and counts a recipe only when every ingredient is on the shelf.
The finder covers recipes with up to 16 ingredients from Food.com, Allrecipes, Epicurious, MyRecipes, Cookpad, Taste of Home, Food Network, Food52, Food & Wine, Serious Eats, Delish and Land O'Lakes, because the other sources in the collection, including cookbooks.com, are no longer online. The collection leans heavily on American websites, so all of this describes American home cooking as it is written online.
- Bień M, Gilski M, Maciejewska M, Taisner W, Wiśniewski D, Ławrynowicz A. RecipeNLG: A Cooking Recipes Dataset for Semi-Structured Text Generation. INLG 2020.
- Eat Your Books, Why is 350 degrees such a common oven temperature? (2021).