Search engines are pretty good at matching keywords with relevant websites. Xiao Li is helping teach Bing, Microsoft’s search engine, to go a step further: figure out the specific task a user is trying to tackle with a query, whether it’s buying a digital camera or booking a hotel room, and return the most useful results related to that task.
Li created software that can automatically crunch through terabytes of Bing’s logs. By building relationships between keywords, the links people click, and the type of information presented on Web pages, the software can predict what a user is trying to accomplish, even if unfamiliar search terms are used. Once Bing determines the intent of a query, it can pass the query to one of a number of specialized search engines that index the most relevant subsets of the Web and can offer task-specific tools. For example, if Li’s system decides that a user searching for “pulled pork downtown” is more likely to be trying to make a dinner reservation than to find a butcher shop, it can kick the query over to a specialized engine that deals in restaurant searches, providing quick links to reviews and reservation systems. —Jessica Mintz