Cornell University researchers are developing search engine software that can learn by noticing which links are clicked on in a list of search results, and how searches are reworded when the first results are unsatisfactory. The research will lead to methods that improve search quality, especially on specialized Web sites such as scientific or legal collections. "I think there is a potential for commercial impact, improving quality and productivity," says Cornell professor Thorsten Joachims.
Joachims and fellow Cornell professor Robert Kleinberg have created an open source search engine called Osmot, which uses extensive machine learning technology. With the help of Cornell professor Geri Gay, eye-tracking studies have shown that search engines can improve by shuffling the order in which results are returned, since a result that shows up toward the bottom of a list may not be clicked on because the user does not scroll down far enough.
Joachims says search engine results represent a tradeoff between displaying the best ranking based on existing data, and experimenting to be able to provide better results in subsequent searches. "The key is to balance the tradeoff between presentation and experimentation in an optimal way," he says.
From Cornell Chronicle
View Full Article
Abstracts Copyright © 2010 Information Inc., Bethesda, Maryland, USA
No entries found