Watson Offers Hands-Free Search of the Web and Beyond

In the search for the next big thing in search, a number of developers are poking around in contextual search, designing software applications that can look at the document on your computer screen, figure out its meaning and significance, and then go get other documents that are related to it but that you might not have known about—or remembered, if those documents are on your own network or desktop.

That’s the space Chicago-based Intellext has chosen to play in with their Watson application. Watson is a downloadable application that sits on your computer, either in a licensed form for the enterprise version or an ad-supported one for the free software, and takes note of what you’re working on, then goes out to the Internet, or to your Word, e-mail or contacts files and gets the relevant articles without being asked, serving them up in a sidebar.

Watson had its origins in the 1996 Chicago Democratic convention, of all places, where now-Intellext CTO Jay Budzik and Kristian Hammond, his former computer science professor from Northwestern University, were managing computer systems for the delegates. They noticed that many questions asked of the system were poorly phrased and began thinking about the way machines derive meaning from words on a page.

That led to an application that could pick out the topics in an open document, use artificial intelligence to determine their proper contexts and then go to search indexes such as Google, Yahoo! and MSN Search to find appropriate references.

Watson began life as an enterprise search application, tested with early investor Motorola and then bundled with business-apps packages from companies like Microsoft. But when Intellext got the opportunity to develop a Watson plug-in for MSN Search last year, they initially offered it as a paid service similar to the enterprise product– $10 a month, or $99 for a year, with a 30-day free trial.

“The overwhelming feedback we got from general MSN users was, ‘We don’t pay for search—search is free,’” Budzik says. “So we decided the product was at a development point where we could let people just download it for free over the Web and we could make money though advertising.” In late March, Intellext signed a deal that will put pay-per-click and pay-per-call ads from the Miva contextual network on top of the Watson display—one ad, periodically refreshed depending on the content Watson is pointed at.

When users work on a document or open a Web page, Watson displays relevant Web content under a series of tabs with “Top Results”, first, then Web pages, news, blogs, premium content, shopping and desktop search results. The tabs have item counts so users can see which categories hold the most results. Once the tab is open, users can see the headlines for the Web pages or desktop documents Watson has found, together with a few lines of description and the source of the item; rolling the mouse over the result shows a longer description; and clicking on the result opens the Web page or document directly. Users can also highlight words or phrases in a document they have open and ask Watson to refine its search based on those highlights.

How relevant are the results Watson finds on its own? Better than the ones users can find by entering a keyword into a general search engine, according to the company. Intellext says that development testing showed that independent expert searchers found nine out of 10 documents found by Watson to be relevant to a document, compared to only five out of 10 when those searchers went on their own to the same engines Watson uses. About a week’s worth of informal testing found that the results are generally very targeted; writing this article with Watson running in the background produced only one irrelevant clinker in a set of 57 Web results—a page about Disney word search puzzle book.

Watson resides on the desktop and doesn’t send any information about documents or their contents anywhere off the PC, so security fears are minimal. The application can work with any document longer than about a paragraph of text and is by default “always on”, although users can change that setting, particularly if they find Watson’s continuous refreshing of its search results too distracting. They may also have to decide if they want to give up the screen real-estate required by keeping the Watson application open. If not, it can be reduced to a toolbar icon that flashes when new content is found.

The way Intellext has defined its own user market keeps it from coming into direct competition with any of the big search providers, even now that they have all deployed desktop search tools. “We don’t have to compete with Google’s index size, and we don’t have to build our own Internet search engine like MSN,” says Budzik. “General search engines today are very good at what they do: indexing the Web and answering queries. But you have to go find them and get them to do that. Proactive search is about getting more value out of those existing systems.”

That position also differentiates Watson from Pico, the “latent search” tool offered as a free download by Blinkx earlier this year. Pico also goes out looking for content without being asked; but it relies on Blinkx’s proprietary search engine to scour the Internet. Truth be told, Pico also looks actively for multimedia content such as video and podcasts related to open documents—very useful for scouting out all that consumer-generated content that’s bulking up the Web.

But Budzik says that with Watson’s open architecture, it won’t be difficult at some future date to integrate the application with one of the other specialized audio and video search engines on the Internet. “That’s what we did when we wanted to include shopping functionality in the free consumer version,” he says. “We simply went out and integrated with eBay and Amazon. It wasn’t any heroic effort.”

“It was a huge decision for us very early on that we would never build a search engine but would use open architecture,” says Hammond, one of Intellext’s co-founders and its chief scientific adviser. “We decided we’re smart enough, but we’re no match for an army of smart guys. And it has turned out that the functionality we can provide is unbelievably flexible.”

That decision to overlay on others’ Web-indexing refinements may make this Watson’s moment, Intellext believes. With both engines and users becoming more sophisticated and thorough in their efforts to search for content, the time is right for an application that both simplifies that search and broadens it.

“Right now, both in desktop and Web-based search, there are fist-fights and massive competition going on among Google, Yahoo!, MSN, Ask,” says Hammond, an Intellext co-founder and the company’s scientific advisor. “We just rise above that, because we can hook to anything. We want Google and Yahoo! to be as powerful as they can get, because the better their indexing performs, the better we perform. We wind up being the connective tissue between whatever you’re working on now and the entire world.”