Led by Neil Roseman, Evri Wants to Understand Content of Every Web Page (and Connections Between Them)

First of all, please don’t call it a search company. Not even a “semantic” search company (one that uses natural-language processing of text), which is what its technology is based on. Evri, a Seattle startup backed by Paul Allen’s Vulcan Capital, is about more than search—it’s about browsing, understanding content, and “connecting the rest of the Web to the things you care about,” according to founder and CEO Neil Roseman. Plus, he said, “We have the most beautiful space of any startup in Seattle, and we’re hiring.”

Looking around at the company’s digs, it was hard to argue with him. A wide-open loft space in Pioneer Square with huge, sunny windows (at least in summer), wood floors, and a ping-pong table covered by a lavish spread of hors d’oeuvres and chocolates. I was there yesterday afternoon to attend the latest Seattle Lunch 2.0 networking event, organized by Josh Maher, whom I met last week at WTIA’s summer party. Roseman, a former VP of technology at Amazon, gave a demo of Evri’s beta site (up and running for a month), which I signed up for earlier this week.

As I understand it, the basic idea is to own the space between search, which Roseman says makes up only 5 percent of all page views, and content, which makes up 50 percent. “It’s called a Web browser, not a Web searcher,” he says. The point of Evri is to get more users to browse the Web continuously, jumping from article to article, Wiki to Wiki, without having to stop and do a bunch of separate searches for relevant information. Evri isn’t looking to replace Google or other search engines, but it hopes to cash in on another way of driving people to their destinations on the Web.

To that end, Evri’s software analyzes content on the Web—news articles, blogs, Wikis, even videos and podcasts—and helps you find related pages about what you’re interested in. If you sign up for the beta, you can open up any article about Barack Obama, say, and Evri’s widget gives a list of recommended reading, organized by topics suggested by the specific article, and also shows you a graphical picture of which topics Obama is most strongly connected to at the moment—such as Iraq, Berlin, Europe, New Yorker, or Michelle Obama. More broadly, it also shows you what the most popular people, places, and things are on the Web (or its sample of the Web) at the moment.

It does all this through natural-language-processing algorithms that analyze the semantic structure and meaning of text on the Web. “Most of how we understand the world is in terms of subjects, verbs, and objects,” says Roseman. The core technology comes from a division of Seattle-based Insightful that was bought by Vulcan. Evri’s software crawls websites, monitors news feeds, and does statistical analysis of content, updated every 10 minutes. It’s all done automatically, without a large team of editors or people interpreting text, says Deep Dhillon, Evri’s director of engineering, formerly of Insightful. He says the core technology is similar to that of Powerset, a natural-language search company recently acquired by Microsoft, but the application is very different.

So where is Evri headed? The company has about 35 people now, and has received some $8 million in investment from Vulcan, including the intellectual property and talent from Insightful. In the next few weeks, it will roll out a new site with a smoother user interface and more features, such as a “find” box (not “search”) that you can type in that connects to pages indexed by Evri. Eventually the Evri network could be supported by advertising. The goal is “not to be a destination, but to be ubiquitous for people interested in how things are related,” says Roseman. “Step 1 was to find a great team. Step 2 is to build a great product. Step 3 is to make a lot of money.”

Trending on Xconomy