Whisk, the U.K.-based service for matching online recipes with online ingredients-shopping, went live in a big way at year’s end, with a partnership with TV channel and recipe publisher Food Network. As its iOS and Android apps rolled out to accompany its browser plug-in, Food Network in the U.K. featured a button on its recipe search engine for a widget that taps into the service, which is underpinned by semantic technology and a cloud infrastructure. A recent second round of angel funding also has taken the service’s total investment to more than £500,000.
Whisk co-founder Craig Edmunds reports about 12,000 app downloads so far, and about a 1.5 percent steady click-through from the button on the publisher’s site – right where it expected to be at this point, he says. Getting the big-name Food Network signed on actually changed plans a bit for the service, which The Semantic Web Blog covered earlier here, and whose co-founder Nick Holzherr was a keynote speaker at the London SemTech event.
Whisk had been planning on a soft launch with a few small publications, but “being on a massive site like Food Network meant the quality of our semantic understanding had to be significantly higher,” Edmunds says. The service performs natural language processing in real-time to translate a recipe to its internal language and from there to that of online markets. “So we brought in a hybrid workflow, basically augmenting the automated semantic analysis with some manual help. Things don’t go live without someone having signed off that a recipe is good.”
Looking ahead to this year, Edmunds’ expectations for the second quarter timeframe are for the service to come to the American market with a U.S. retailer and recipe site publisher. The intent, he says, is grow the relationship with Food Network, but no partners have been confirmed as publishers or retailers stateside as of yet. “There’s a lot of backend work we have identified about what are the semantics the system needs to extract from the data that is passing through, and how to refine the model to even better represent the underlying data,” he says, and notes that some “really cool features” are also on the way.
Whisk will look to discover what else it can tell users about the exact groceries they’re buying – dietary impact, nutritional information, even what wines will go with a particular recipe. “The core of what we do is use semantic technologies to understand recipes, and once we’ve identified those semantic units within the recipe, we loop that into all the other information, like ingredients,” Edmunds says. “We have a lot of intrinsic knowledge of all those different structures and the relationships between them, and that lets us extrapolate these sorts of features.”
Take wine as a complement to food. Most non-experts follow the old rule of white meat, white wine, red meat, red wine. “But actually there are flavor combinations in a recipe that create a taste profile, and using that taste profile you can select the perfect bottle of wine, or certainly the region, grape variety, and country of origin [that would pair best with the food] – all those sorts of things,” he says. “They map directly to the concepts of the recipe.”
Also being looked at is the idea of making it simple for people to collaborate with friends and family on weekly online grocery shopping, which Edmunds says is more developed in the U.K. to date, with all five major grocery chains there onboard. He sees opportunities for Whisk to make it easy for spouses or housemates who see a recipe they want to cook to send that to the house grocery shopper’s list so that those ingredients can be merged into the online order, too.
There’s also the idea of the connected kitchen, forging a path from the recipe added from a web browser to the ingredients list kept on an iPhone to a tablet app that comes into play at cooking time. Whisk, in fact, is already working on ideas for the tablet end of things. “You want to be able to pull up recipes that you recently purchased [ingredients for], maybe drop in a video from the recipe publisher with a celebrity chef cooking the dish,” Edmonds says, and watch it online in the kitchen on your tablet as you’re getting ready to prepare the meal. Perhaps the app might even interact with the new age of smart ovens, to set it to the proper temperature, for instance.
With tens of millions of impressions a month already, Edmonds says the cloud architecture Whisk leverages for the service still has plenty of headroom as it expands to U.S. shores. “We are at the production grade in terms of our infrastructure,” he says. There might be a bit of a test phase with any new publisher but things are otherwise in a ready-set-go state. “Now it’s just get the deals done, update the models slightly for the U.S. market and get data,” he says.