Social network of food picks best Thanksgiving recipe
Posted: Sun Nov 25, 2012 8:10 am
Hal Hodson, technology reporter
Unleash big data on that Thanksgiving turkey! For anyone preparing a feast today, the hidden connections between your foodstuffs probably aren't your first concern.
But they do exist, and it turns out that they play a large part in determining how well-liked any given recipe will be. Lada Adamic from the University of Michigan downloaded 46,337 recipes from the website Allrecipes.com, then mined the text for ingredients used, the type of meal, the kind of event and the region where the recipe originated. Combining these data with the reviews each recipient received, Adamic built a model that can predict - with around 70 per cent accuracy - how many stars a recipe would receive based solely on its ingredients. This should help you put together a recipe that will unthrone the current top-ranked Thanksgiving meal on the site - awesome sausage, apple and cranberry stuffing.
Adamic's network analysis is similar to the kind which lets computer scientists understand where slang words originate, or which Romantic novels are the most influential.
She also built a second network that showed which ingredients are most interchangeable, by analysing the alternatives for each recipe suggested in the millions of reviews. Apple sauce for vegetable oil is a popular one, for example.
Writing on her personal blog, Adamic noted the most popular Thanksgiving substitutions, according to her Allrecipies.com model. "Cranberries end up substituting for other kinds of fruits and even somehow for chocolate. In the fat category, olive oil and butter seem to be recommended as substitutes for things such as margarine. Yams are often recommended as a substitute for sweet potatoes (more so than the other way around)," she writes.
University of Michigan PhD student Edwin Teng even went as far as to whip up a Thanksgiving- specific version of the network. Unsurprisingly, turkey, cranberry and pumpkin all feature prominently. Tuck in!
- (Image: Robert Linton/Getty)
- 155437421[1].jpg (82.27 KiB) Viewed 2656 times
But they do exist, and it turns out that they play a large part in determining how well-liked any given recipe will be. Lada Adamic from the University of Michigan downloaded 46,337 recipes from the website Allrecipes.com, then mined the text for ingredients used, the type of meal, the kind of event and the region where the recipe originated. Combining these data with the reviews each recipient received, Adamic built a model that can predict - with around 70 per cent accuracy - how many stars a recipe would receive based solely on its ingredients. This should help you put together a recipe that will unthrone the current top-ranked Thanksgiving meal on the site - awesome sausage, apple and cranberry stuffing.
Adamic's network analysis is similar to the kind which lets computer scientists understand where slang words originate, or which Romantic novels are the most influential.
She also built a second network that showed which ingredients are most interchangeable, by analysing the alternatives for each recipe suggested in the millions of reviews. Apple sauce for vegetable oil is a popular one, for example.
Writing on her personal blog, Adamic noted the most popular Thanksgiving substitutions, according to her Allrecipies.com model. "Cranberries end up substituting for other kinds of fruits and even somehow for chocolate. In the fat category, olive oil and butter seem to be recommended as substitutes for things such as margarine. Yams are often recommended as a substitute for sweet potatoes (more so than the other way around)," she writes.
University of Michigan PhD student Edwin Teng even went as far as to whip up a Thanksgiving- specific version of the network. Unsurprisingly, turkey, cranberry and pumpkin all feature prominently. Tuck in!