How Semantic and Social Search Are Evolving: Lessons From the Evri-Twine Merger

3/15/10Follow @gthuang

What’s really happening in the emerging fields of semantic and social Web search? I’ve been thinking about this since last week’s merger of Seattle-based Evri and San Francisco-based Radar Networks, the developer of Twine.com. The real story is not that one Paul Allen-backed company has acquired another, or that any investors got hosed in the deal, but rather that semantic and social search are converging in a complex way—and that giants like Google, Microsoft, and Amazon are paying close attention to these developments as they work to stay ahead of nimble startups.

The semantic search sector is where researchers and companies are using advanced algorithms to better understand what people mean when they search for content on the Web. Semantic search is also about making connections between online entities like people, places, and products, so as to present results to consumers in a more useful way. These techniques use natural language processing and other technologies to try to go far beyond the traditional methods of Web search, like matching keywords and ranking the relevance of documents. But they must work well, and small efforts have struggled to get enough traffic and user data.

An example might be that if you typed in “hurt locker seattle,” you would get back information about “The Hurt Locker,” local show times and locations, and also links to news about the movie’s cast and crew, their related projects, and other war-themed movies—all in a way that reflected your personal browsing interests.

Meanwhile, social search, which is sometimes called real-time search, has made more progress in the journey from concept to practical reality. You can think of this as search engines that incorporate up-to-the-minute data from social sites like Twitter and Facebook. Google and Microsoft’s Bing already have strong efforts in this arena, and many startups are actively building search engines for Twitter and other social media.

It’s clear that Web search is going through some major growing pains. And as search technologies become more integrated into everyday activities like shopping, social networking, traveling, and mobile Web browsing, more companies across a wide spectrum will have a stake in it. In other words, Amazon, eBay, Facebook, Twitter, Expedia, Kayak, and Apple’s iPhone app store all have social-semantic search problems. A key question for entrepreneurs and investors is, what is the best approach for building and marketing a startup in this sector? I’ve gathered a few perspectives from outside experts to help focus the discussion.

But first, some more details on last week’s merger. Will Hunsinger, Evri’s CEO, tells me he started talking with Radar Networks founder Nova Spivack last year about working together, and that they saw a cultural fit between their companies, which were solving “similar consumer problems.” Hunsinger says the combined organization of just over 30 staff (about 10 from Twine) is looking to consolidate its efforts into one website in the coming months. He also says Evri is “not walking away from” the media partnership aspect of its business; that includes deals with organizations like Hearst, The Washington Post, and The Times of London.

But the company’s central vision, he says, is to deliver “intelligent [data] streams to consumers” that, in effect, reduce their need to do traditional Web searches. That means providing them with a steady stream of personalized news and information gleaned from the Web. Instead of tracking five or six categories on Google News, you could create thousands of categories and receive feeds about any topic, however specific—in sports, you might get streams about major league baseball pitchers, football players with Super Bowl wins, or college basketball teams in the NCAA tournament. It could be like an automated, smarter version of Twitter—or it could actually incorporate Twitter, helping you sort tweets by meaning or category.

Evri says its intellectual property position is particularly strong after gaining some key patents from Radar Networks, including ones on natural language processing and semantic understanding of text. (Twine was less focused on real-time streams, and more on learning about people’s interests.) The merged company’s patent portfolio now includes more than 15 issued patents and 40-some others pending review. Hunsinger says the merger “makes us the lead dog in the semantic search and discovery space.” He adds, “Now we have to figure out how one plus one equals three.”

Indeed, it will be interesting to see how Evri makes use of the hard lessons Radar Networks learned over the past couple of years. Many of these lessons were detailed in an extensive blog post by Radar Networks’ Spivack last week. They include nuggets like his advice to early-stage entrepreneurs about raising less money from VCs, spending less, getting to profitability quickly, and not staying on as the long-term CEO. All in all, it doesn’t sound very encouraging, and it’s enough to make anyone wonder whether semantic search is still a solution looking for a problem. But perhaps handling social, real-time search in a better way could be that problem.

“This will certainly splash some cold water on investors,” says Steve Judkins, founder of Seattle-based MemeSpring, an early-stage startup focused on semantic search. Still, he remains bullish on the potential for innovating in the sector. “I believe the opportunities from social-semantic search startups remain as big as ever, but only if the business model makes sense from a very early stage,” he says.

That means getting the model right at a small scale and then increasing the technology investment, Judkins says. The key, he says, is to develop on the cheap and “don’t wave the red flag in front of Google.” In other words, realize that as a startup, you won’t beat the big search engines at most applications, and that you need a clear plan for getting traction with a small number of consumers by focusing on a particular niche area—like social media tracking, contextual discovery, or semantic advertising.

But where semantic and social search are headed still isn’t clear. “Overall it is obviously early days in the space. There is a lot of activity from startups and from Google and Microsoft, but the category hasn’t really been defined,” says Cameron Myhrvold, a founding partner of Ignition Partners, based in Bellevue, WA. Myhrvold leads Ignition’s investment in Topsy, a social search startup in San Francisco focused on Twitter, around which he is seeing a lot of activity. “Growth is immense, and at least at Topsy, user volumes and data volumes are exploding,” he says.

Of course, even if a startup’s business model works and it gains some traction, its product might not be better than what the incumbents will provide. Brian Bershad, Google’s Seattle engineering and site director, said in November at an Xconomy event held at the University of Washington, “I would actually not encourage small companies to go after anything in search having to do with text, because I think we’re going to get there.” Other experts at the event suggested that startups would be smart to pursue search in other niche markets like video or location-based services.

That doesn’t really help Evri, though. For better or worse, the company doesn’t seem to be in a huge rush to get to profitability. “Right now we’re really focused on building out the product and getting the right experience,” Hunsinger says, but he is also “thinking about the core revenue model.” Although he wouldn’t share any details just yet, this is certainly crucial to the near and long-term future of the company—and the semantic search landscape. Paul Allen’s Vulcan Capital may have deep pockets, but it is doubling down on Evri, and it needs to see some results.

And that dovetails with what Evri’s front man said when I asked him about the most challenging aspect of his job. “This is a highly competitive, fast-moving space to be in,” Hunsinger says. “You don’t have the opportunity to noodle too long on where you’re going to stake your claim.”

Gregory T. Huang is Xconomy's Deputy Editor, National IT Editor, and the Editor of Xconomy Boston. You can e-mail him at gthuang@xconomy.com. Follow @gthuang

By posting a comment, you agree to our terms and conditions.