You are here:  Home  >  Conference and Webinar Communities  >  Current Article

Real-Time NLP And The Cloud Are Key To Online Recipe And Shopping Service Whisk

By   /  September 5, 2012  /  No Comments

What do you get when you mix two parts natural language processing with a little personalization, and add in a dash of the cloud? The answer is Whisk, a U.K. company building a service that lets users purchase the ingredients for any recipe they find on the Internet.

“The crux of it is that you can take any recipe on the ‘Net and turn it into a transaction in on online market,” says co-founder Craig Edmunds. “There’s a machine translation problem from the recipe up through to our internal language, which is one NLP problem, and then another is from our internal language into online markets.” Another leg of the work is that the service seeks to not match to just one item at a market but as many as possible, and consider user preferences as to which is the optimal product, too.

At the upcoming Semantic Technology and Business Conference in the U.K., Edmunds will be considering how the issues of machine translation, manual intervention, personalization and the cloud intersect in creating a service that adds all the ingredients they need for dishes they find online straight into their online shopping basket.

In particular his speech to attendees will hit on the issue of doing NLP in a real-time environment: “In a lot of problem domains it’s acceptable to run NLP analysis in offline batch overnight, or more or less frequently,” he says. “But our whole point is to take any recipe, and to translate it you must be able to do it immediately. You can’t tell people to come back tomorrow to buy an ingredient.” It must be made available for transactions right away.

To help solve the challenge, Whisk built its software stack to be optimized for scaling, with the eight main components of its application architecture able to be run and scaled individually. “So, if we found a lot of web pages are being sent in that we never saw before, we can scale up a whole load of web scrapers to look at those pages and bring them back to turn into our internal data structures. And if we found a load of NLP problems for recipes, we can just scale up and add extra nodes on that piece of the infrastructure, and that applies all across our system,” Edmunds says.

Obviously, it’s the cloud that makes this possible. Whisk works with Microsoft’s Azure cloud platform, which Edmunds says provides startups good flexibility on costs and free reign over the number of services needed.

“How we designed the solution and made it compatible with the cloud so this can happen in real-time is what we think is the interesting bit,” he says. “Having the cloud, the infrastructure-as-a-service providers, where you can provision almost an infinite amount of resources for a very short period of time completely changes the way that things are calculated.”

But the service does come across some interesting issues that require human intervention more than cloud scalability. As well as helping to calculate what you need for the number of portions you’ll be serving, the service also aims to work out recipes to cook with leftover ingredients. For all this, it has to understand comments in recipes like adding a pinch of this or a handful of that. “There is no exact way to quantify how that maps to store items. A few times we have had to literally figure out what a pinch of salt is, sitting there with weighing scales with a bag of salt and then pinching it to see how much it weighs, or a handful of coriander and figure out what that is. Then we store all that translation logic within our system.”

Part of the semantic understanding the service offers is also understanding the consumer and his or her social graph and demographic information to help personalize item choices for them, he says. So, for the working-class family without a lot of disposable income, it might be more useful to suggest the basic brands, whereas those with higher incomes might prefer to have premium range product choices. “We are trying to figure out all that information, ideally with as little user intervention as possible, and then we are continually learning as the system is used,” he says.

The service, which will initially be available in the U.K. only, is still being built up, but the team is demoing it to recipe publishers it might like to partner with. The goal is for the app to be fully launched by month’s end.

You can register to attend the London SemTech event here.

About the author

Jennifer Zaino is a New York-based freelance writer specializing in business and technology journalism. She has been an executive editor at leading technology publications, including InformationWeek, where she spearheaded an award-winning news section, and Network Computing, where she helped develop online content strategies including review exclusives and analyst reports. Her freelance credentials include being a regular contributor of original content to The Semantic Web Blog; acting as a contributing writer to RFID Journal; and serving as executive editor at the Smart Architect Smart Enterprise Exchange group. Her work also has appeared in publications and on web sites including EdTech (K-12 and Higher Ed), Ingram Micro Channel Advisor, The CMO Site, and Federal Computer Week.

You might also like...

Three Traditional Storytelling Techniques That Add Value to Data and Analytics

Read More →