Blog: Google Schools Its Algorithm

Google Schools Its Algorithm
New York Times (03/05/11) Steve Lohr

Researchers at Google and elsewhere are straddling the forefront of computer intelligence through their efforts to program machines to understand human language. Improvements to statistical algorithms are possible due to an ever-increasing corpus of language on the Web and the development of faster computers, but machines face a sizable challenge with parsing and categorizing language. Google's Web site-ranking algorithm has a heavy reliance on connecting search terms to noun phrases in a Web page, as well as a site's popularity and how frequently other sites link to it. Google recently announced that it was performing a major retool of its ranking formula to downgrade low-quality sites that are mainly set up to siphon traffic from Google's search engine. "As we improve the language understanding of the algorithm, all the cheap tricks that people do will be recognized as cheap tricks instead of tricks that work," says Google fellow Amit Singhal. Computer scientists say the future of search engines such as Google resides in leveraging innovations in machine learning and language processing to transform them into answering machines.

View Full Article