Is your search engine failing to serve up the pages you want? Do you wish it could understand your hobbies as well as you do? The solution may be on the horizon, in the shape of a search engine that automatically ranks results according to information about your personal interests it gathers from your PC.
Personalising searches has long been a challenge. Search engine Vivisimo allows searchers to personalise pages manually by clustering the results of ambiguous searches. This allows subsequent searches for 鈥渏aguar鈥�, for example, to group pages into separate categories for cats and cars.
Now Jaime Teevan of the Massachusetts Institute of Technology is trying to get rid of the first, manual stage. Along with Susan Dumais and Eric Horvitz of Microsoft Research in Redmond, Washington, she has created a system that sits on a user鈥檚 desktop, indexes the emails, documents, calendar items and web pages, and deduces their particular interests. It then automatically reorders results served up by a search engine according to the words it has found from its desktop search.
Advertisement
Teevan has tested the system against the unaltered results served up by MSN Search, with partial success. 鈥淪ome queries personalise very well, and others do not,鈥� she says. More work is needed to identify which queries work best. The system was presented at an information retrieval conference in Salvador, Brazil, last week.
In future, the user鈥檚 location could be taken into account if they were using a GPS-enabled laptop or phone.
