So That’s What Wavii is Up To: “Making Facebook out of Google”

11/30/11Follow @curtwoodward

Seattle startup Wavii, previously seen taking the concept of “stealth mode” to extreme lengths, is letting slip more details about the product it’s been building for these many quiet months. In a pair of blog posts today, the startup says it’s using machine learning to sort content from across the Web into feeds—as they say here, “Making Facebook out of Google.”

Yes, a variation of the well-worn startup pitch “it’s like X for Y.” To be fair, that’s not a bad way of relating a bunch of really nerdy computer science stuff to garden-variety online content consumers, which looks like the target audience.

Wavii aims to sort through the mountains of information online and present related pieces of content together in a stream, as if there were a person curating the feed. That’s what Facebook (or Twitter) users do for their followers: Call pieces of information out of the ether and pass them along to others.

Wavii’s reference to Google points to the fact that, while they’re very good at finding information, search engines still rely on pretty specific queries that hit the right keywords in more or less the right order. Commercial search isn’t so great at answering open-ended or natural-sounding questions, like “what’s the traffic like on I-5 right now?

That’s one of the things University of Washington search expert Oren Etzioni has been studying, and kind of harping on, in his spare time. As I wrote in this profile of Decide, the e-commerce startup that Etzioni co-founded, a technology called open information extraction actually does allow computers to comb through written materials and decide whether the content is, in fact, relevant to a specific topic.

Decide uses open information extraction to comb through product announcements, press releases, blog posts, and news reports to figure out whether a newer model of a particular electronics gadget is coming out soon. If you wanted to do that yourself, you’d have to Google a string of keywords and scroll through several pages of results to find the right information. Decide’s machines can do it automatically and continuously by parsing the written material already out there.

It looks like Wavii is using a similar approach to curate content about pretty much any topic a user might want to follow. In today’s post from someone identified only as Dan (stealth mode), Wavii labels its version of this technology “big data machine learning.”

“The key to making this technology work is teaching the computer to recognize the things we care about when they’re being talked about, even when they are expressed in tens or sometimes even hundreds of different ways, all without making too many mistakes. Once the computer has this, it can easily build the feed items you’re used to,” the company wrote.

Wavii hinted strongly at this approach in an earlier guest post on TechCrunch, using the wedding of Prince William as an example. Former TechCrunch editor Michael Arrington also reported some of these concepts in an earlier post about Wavii, based on a test version he saw. So I guess we can officially declare an end to this particular tour through “stealth mode,” or as GeekWire’s John Cook recently put it, “double secret stealth mode.”

I e-mailed CEO Adrian Aoun to see if he could tell me any more about Wavii at this point. At last report, from Arrington in July 2010, the company had raised about $2 million from various angel investors.

Curt Woodward is a senior editor for Xconomy based in Boston. Email: Follow @curtwoodward

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