Posted by: Rob Hof on October 1, 2007
We’re all so trained to typing a couple of keywords into Google that we forget there are other ways of searching for and finding stuff online. A few upstarts such as Powerset and Hakia are trying to improve search results by analyzing the meaning of words and the semantic context of Web pages to let people use more natural language. Now, another upstart is trying a completely different method.
Munich-based Proximic, whose investors include Wellington Partners, backer of the fast-rising European social network Xing, and Holtzbrinck Group, publisher of Scientific American, calls its new search technology “pattern proximity.” Instead of matching keywords or trying to divine meaning of words, it analyzes patterns of characters and character sets within documents. That means it’s potentially independent of a particular language. There’s more detail on this here, and Robert Scoble has a video with CEO Philipp Pieper and CTO Thomas Nitsche. “I’m intrigued,” says IDC analyst Sue Feldman. “Keyword matching has definite limitations.”
The technology sounds too good to be true, though in trying out various early examples (click on the "related" link), it seems to bring up some useful and very specific related links. (Unfortunately, the Firefox browser add-on that aids in finding related content didn't like my PC or something, and I was forced by the add-on or some other completely coincidental problem to reinstall Firefox.) Proximic seems to go well beyond somewhat related services such as Sphere and Snap in that there's new search technology behind it.
Proximic claims its patent-pending technology, informed by Nitsche's experience as 1984 world computer chess champion, requires so little memory and computing power that even in indexing the entire Web through its network of distributed servers, it may never need to build a huge data centers. Hard to believe, but we'll see as Proximic gets rolling.
The technology is just one part of what Proximic is going public with today. It's also launching what it calls the first global "contextual content network" that will let publishers point people to related, very specific stories that general-purpose search engines might not show results for because they're not linked to by many other Web sites. So far, the U.K. daily newspaper The Independent and Nature Publishing Group have joined the network.
Of course, better search results mean better ad targeting, and that's the ultimate appeal from a business standpoint. "The interesting thing is the impact on ads," says Feldman. "The chances of increasing the click-through rate are much higher." Ian Mulvany, product development manager for an early user, Nature Publishing's social bookmarking service connotea.org, says Proximic not only provides more controlled search results but better ad results too. Google, he says, kept showing ads for "tea," even though his site obviously has nothing to do with the beverage.
It's hard to tell whether an unknown upstart's technology and ad network prove out, especially at Web scale. So it seems unlikely that Proximic by itself will challenge the likes of Google directly--even though as Pieper points out, "We're competing for real estate to present advertising." More likely, it will license its technology to various e-commerce companies such as Amazon.com or eBay, ad networks, or a portal such as Yahoo!, or get purchased outright. Either way, it's pretty clear that Google's dominance isn't deterring competitors from trying out new things. And that's going to make search an interesting contest for years to come.