Thursday, September 29, 2005

A Portal in Reverse: Watson Puts Content from Internal and External Sources in Context

I sometimes think of the Microsoft Research Task Pane and wonder what it would have been like if Redmond hadn't been trying to prop up its Office Suite as the core objective of this somewhat ill-fated tool. The answer comes in some degree in the form of Watson 2.0, a desktop search tool from Intellext that does a nifty turn on getting content in the context of what matters most to us at a given moment. Watson is a piece of software that sits on a PC and can find content that matches whatever one is viewing in common desktop applications such as a browser or word processing. Watson uses the current desktop document as "training data" to target searches on content residing on local drives (compatible with Google Desktop), the open Web, premium content from HighBeam and other sources, including sources on an intranet when it's configured for enterprise use. The Watson tool can sit as a sidebar next to your main desktop area or be undocked for background use.

On a 4-page proposal I was working on in Word I fired up Watson and got an interesting and pretty relevant range of search results in the "Top Results" window and good results in some of the other results folders for desktop and other content sources. Once you've found Watson results for a document they pop up in the Watson tool every time that you display the document - instant context. Conceptually this is exactly how the "portal" of the future is going to work: not forcing users to "go" to a Web site and to ask for things that might match a need and that may or may not fit a user's workflow but instead for the library of content that's most relevant to a given task or focus to be always ready and refreshed at our fingertips, like a bookshelf that changes with every document that I pick up. It's also an excellent way to have local resources instantly available to those who love to use a Web search tool as a "go-to resource: users can still do that but their other references will be right there along with the search results.

But as with any search tool using training data, the focus of results is only as good as the source being used. Using smaller documents may yield less perfect results. I tried opening up a blank Word document and typed in the beginning of a familiar story: "Once upon a time there were three bears: a momma bear, a poppa bear and a baby bear." Instead of Goldilocks-style results I got a lot of documents about real-world bears and even the home page of Bear, Stearns, an investment bank. Using a Web page for the story "The Story of Goldilocks and the Three Bears" as Watson input brought back similarly general ursine results. Yet a search on Google without Watson using the "Once upon a time" sentence brings up some pretty accurate fairy tale results. So for some people used to accurate searches with short phrases or keywords Watson may not be the ideal tool.

Watson is a well-designed tool that offers a lot of useful context for content resources and seems to shine best with relatively finite desktop collections; presumably it would do fairly well with enterprise resources also. With PC screens becoming wider by the day there's going to be plenty of screen real estate for this kind of sidebar tool in many environments. For companies trying to pull together internal resources for "Google-ish" users Watson offers an interesting alternative to lengthy and expensive portal rollouts that can fit in easily with established work habits. Tools like this may be easily absorbed as features into wider offerings, but for the time being Watson offers a good way to try out desktop-contextualizing your content for a fairly modest investment.
Post a Comment