Yummly’s semantic recipe search gets spicy

Semantic recipe search engine Yummly is tuning its search features to include a new way to search specifically for recipes that are “spicy” just as it’s hitting a major milestone. After coming out of beta in June 2010, the company says it’s now doing half the unique visitors of Epicurious.com, the popular search engine for Bon Appetit, Gourmet and Self magazine recipes. In July, year-old Yummly reached 2 million unique visitors in a month for the first time, the company is set to announce Wednesday.

Yummly’s game is semantic recipe search, meaning it looks not just for keywords, but context of a search and the intent of a search. One of the ways you can search is using a sliding scale of flavor profiles (salty, sweet, sour, bitter). The site also enables extremely specific recipe searches, say, for a gluten-free strawberry cupcake recipe that calls for a package of instant vanilla pudding.

As it’s grown, people have asked that Yummly expand its categories to be able to search specifically by spiciness, according to CEO and co-founder David Feller. So it did.

It’s a relatively minor update, but adding spiciness–and how exactly they go about allowing users to find something on a sliding scale of spicy–is an example of what the secret sauce behind Yummly does to differentiate itself from a regular recipe search. And it’s another example of how data and food are mixing very well together online, or as my colleague Stacey Higginbotham put it recently, “making food fit for the web.”

“We can deconstruct the recipe and see the elements of it. We look at ingredients and determine if [a recipe] is a particular cuisine, or if it is good for these people with certain diets or allergies,” Feller said in an interview this week. “Since we can understand when we see these ingredients, we can determine relative spiciness of each of these recipes.”

For the data to compare spiciness, Yummly went to the classic rubric for spice, the Scoville scale, which ranks the heat of peppers from bell peppers on up to those even hotter than habanero chilies. Yummly uses that data to rank recipes based on the spicy ingredients as well as add more information not included in that scale, such as mustard, wasabi and horseradish.

That spiciness is Yummly’s first addition to the scope of the search behind the site is somewhat ironic considering the site’s origin story.

Feller, a Silicon Valley and startup veteran by way of eBay/Half.com/StumbleUpon, was inspired to start Yummly because of a notoriously spicy condiment, mustard. He loathes it. So he sought a way to accurately search for recipes without it.

Yummly likes to think of itself as “the Kayak” of recipe search. Its goal is to understand recipes as “more than just text,” in other words, not just to search for keywords. This is where Feller thinks his site has a leg up on other recipe sites and even Google Recipe Finder (s goog), which can–and probably already does–act as a default for many who are simply googling for what to make for dinner.

“Google doesn’t go deep enough to really understand” all the ingredients in a recipe, claims Feller. “It’s still just keyword matching. We go the extra level to connect data and what people are looking for.”

Thumbnail image courtesy of Flickr user VJ_fliks