You are here:  Home  >  Data Blogs | Information From Enterprise Leaders  >  Current Article

Where To Eat? Let Neural Network Computing Help You Decide

By   /  June 26, 2012  /  No Comments

Dollars to donuts most folks haven’t ever found a place to eat courtesy of neural networking technology before. Generally, Internet searches for spots to have a bite come courtesy of friends’ Facebook recommendations, services like Yelp, and even some semantically-powered offerings such as BooRah, now an Intuit company.

But the collection of neuroscientists, computer scientists, astrophysicists, and creative artists behind Nara, launching into public beta today, have taken the advanced neural networking route to automate, personalize and curate web dining experiences for users – though there’s more to come on the future menu. President and CEO Tom Copeman says of the company, which in April secured $3.6 million of a $4.5 million equity offering, that its cutting-edge neural network and proprietary and patented algorithms and process for analyzing tons of web data, and personalizing it, including considering user feedback on the suggestions it offers, is creating a whole new category.

That is the pure-play digital lifestyle brand that “creates an emotional connection between us and the Web. We’re trying to change how people think about the web, and from sense of what it means to me, and makes sense to me, and how personal it is to me.”

Of course, the word personalization has come up in context with numerous services over the years, so why does Nara think what it’s got is special? Nathan Wilson, CTO and research scientist at MIT, from where he also holds a doctorate in brain and cognitive science, a masterʼs in computer science and artificial intelligence, as well as a bachelor’s degree in neurobiology and cognitive, built the Nara neural-network. According to Wilson, the answer to that question is that trying to drive personalization via understanding words or the analysis of the social graph, “is incomplete and flawed. Ours is a differentiating way to do this,” based on accumulating data from anywhere people on the web gather and rate things as well as user feedback, on the intelligence of its algorithm, and its use cases. And, “the right architecture,” he specifies, “is what will define what does and doesn’t work.”

His research at MIT that is behind the technology explores how neural networks in computers link together lots of concepts, much like human brains do, and how the symmetry in the architecture itself pushes to personalization by finding the various and nuanced ways in which restaurants may be like each other. That is, it helps users determine other places they may like to visit based on places they know, but not just by serving up five other Asian restaurants because users’ profiles indicated they liked other Asian restaurants, or showing five other moderately-price restaurants because their profiles indicate that to be characteristic of the usual dining destinations.

Rather, it crunches lots of features and considers things down to more subtle traits, like ambience, or even wine selections, and draws connections among factors that may be very far apart in the neural network. Its implicit tracking of millions of reviews lets it capture and parse patterns, while it also explicitly leverages expert and categorized curation sites, indexing their information about menu items or bottles of wine, for instance, and letting its algorithms do the computational connections across it all. “That where non-neural networks would fail,” he says.

The depth of the connections it makes to show in real-time recommendations of restaurants to users who start their searches by entering three of their favorite restaurants aren’t particularly obvious first-thing in the interface – aside from the fact that some recommendations do share similar cuisines as a favorited mention – but Copeman says that in private beta testing they continually hear from users that the recommendations hit the nail on the head.

“When they would go to those restaurants, they are kind of astounded Nara figured out what would feel right to them and understood them,” he says. For instance, a user may not have entered a Chinese restaurant as a favorite item, but a recommendation for one may occur based on other aspects Nara has figured out are important to the user, like it has a trendy atmosphere and dark lighting. “Starting with simple categories, like restaurants or hotels or wine, those that have clear properties, will get us farther, faster with personalization,” says Wilson.

Copeman points out that the idea is to work for the first user the first time, “We don’t need more than one Nara user to make it work, whereas other sites try to leverage the social layer. We don’t need that, or a lot of Nara users,” he says. Social recommendations are tricky. “The friends you have on Facebook, many of them may be people you went to high school with or work with. There’s too much noise for it to be a reliable signal at this stage. And semantics-based recommendations [are challenging] too, because it’s still so hard to understand words, even though there’s so much important research being done.”

As with Pandora, he says, “there is a lot of discovery and helping to give people a broader perspective so that they don’t have to spend a lifetime searching and reviewing something. Before you know it, they’d starve to death.”

As you may have guessed, the level of computing and scalability requirements behind this requires the resources of a cloud infrastructure, and the service is using Amazon’s to provide its recommendations. Also helping make it possible, Wilson says, is that more information on the web finally is getting more organized and structured.”

Right now Nara serves eight major cities, the idea being to target those living in high-density locales who eat out a lot, and often at the same places, so that they discover new places. And, it also aims to help those visiting a new place for the first time transfer what they like in their native DNA dining in order to deliver appropriate eateries in the new city. If Nara gets its calculations wrong and the user gives a thumbs-down to any experience, the restaurant will come off that user’s future recommendations.

A mobile version is on the way, and at some point an API will be available too so that other sites and categories can tie in to its technology. Nara itself will be expanding into other consumer lifestyle categories but it is entertaining B2B opportunities, too, and other possibilities that won’t cannibalize its own plans.

Solving of the search-to-find problem requires people with traditional expertise in computer and Internet technologies, of course. But Copeman says the other skills the company has accumulated matter a great deal, too – for instance, creative artists are there to deliver the intuitive and emotive front-end to match the powerful and robust back-end, he says, and astrophysicists bring along the big-thinking, Internet-out perspective to help the service grow and scale. Says Copeman, “Nara connects art, science and technology.”.

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 →