intent—why you’re doing the search—based on prior analysis of billions of queries over tens of thousands of categories and hundreds of user scenarios. Then it tries to present more of the relevant information you’re looking for upfront, on the search result page, to help you complete your task without clicking away from the page or starting a new search right away.
Most people have tried the engine already, but at the Microsoft faculty summit on Monday, Bing’s senior director of product marketing, Fred Savoye, walked the audience through some specific examples. Search for “weather” on Bing, and it brings up the local forecast. Search for “Alaska Airlines,” and the company’s official website pops up on top as the suggested “best match.” Search for FedEx or UPS, and the customer service number and package-tracking service pop up. Look up the Seattle restaurant Wild Ginger, and a “get directions” button appears. On any search, the left side of the screen displays common refinements to the query as well as your search history. There are numerous other features, too, like user reviews for product searches, and the “enhanced view” link for certain items, which brings in content from Wikipedia and other sites.
Shum draws a distinction between a search engine, where the goal is to serve up a website and get out of the way as fast as possible, and a decision engine, where the goal is to fulfill the user’s intent and help them explore the results within the search page. In the latter case, a major challenge is how to display the information in the best way. “Now you have the whole page, the big real estate,” he says. “What do you want to present there?”
As he explains, the technology under Bing’s hood boils down to three issues. The first is relevance, the familiar notion of matching a search query to the appropriate URLs. The second is performance—processing the query and sending the results back in a fraction of a second. And the third, which seems to be Bing’s main differentiator, is user experience. That’s everything from the user interface and site design to the analysis of billions of queries and documents for understanding what a particular search query means—and what the user actually wants.
To this end, the Bing team has been developing a “user interaction model.” This is a mathematical approach to understanding how best to narrow down a search query to determine a user’s intent, or help the user expand the query to help it make sense to Bing. By tracking how people use the
