Microsoft's Bing search engine is testing a new technology aimed at tailoring search results to individual user's tastes, two experimental applications that recommends news stories shared on Twitter.
What Google describes as personalization Bing is calling Matchbox, which attempts to make predictions about what searchers might want to see, based on if someone like them has done something with it but also on what the content says.
Traditional personalization technologies from Google use collaborative filtering, which means if user A and user B both like something, the search engine can predict user A's future "likes" based on what user B "likes." It's a sort of taste by association, or in Bing's parlance, collaboration.
Ralf Herbrich, principal development manager for Bing Mobile Personalization, noted that this approach works well for static products, as well as restaurants and movies. However, it does little to help account for evolving news and real-time information.
Herbrich's cites the example of a user reading and rating articles on electric cars. Collaborative filtering techniques would look at other people who have read similar articles and cluster the searcher with them. New articles related to electric cars may be recommended to a user provided a user with similar tastes interacts with it.
But that is a hit and miss value proposition because, as Herbrich noted, it doesn't help the first person searching for fresh content on electric cars find the article.
Matchbox uses the information about the entities mentioned in an article. So because Matchbox knows that Tesla is an electric car, it can retrieve "Tesla Model S: 300 miles on 1 charge" as a viable search result with relying on input from other people who conducted similar searches.
The idea of this "feature generalization" is to grok the Web more like a human than a machine. In short, it's applying artificial intelligence to search.
To test Matchbox in a practical scenario, Bing built Project Emporia a new Web application and Windows Phone 7 app users can test.
Emporia, which Bing demonstrated at South By Southwest Interactive over the weekend, recommends news stories shared on Twitter using the Matchbox technology.
The apps filter news stories by automatically predicted news categories. Integration with Twitter helps users see what stories your friends and friends-of-friends are sharing. Finally, the apps recommend news stories based on searchers' personal preference votes.
"This new approach will help users getting on top of the wealth of news coming at them every day," Herbrich said.
Perhaps, but the technology in its infancy appears limited to social search. Will the technology remain relegated to Web and mobile apps, or make its way into the broader Bing search engine as a regular feature?
More importantly, will it improve Bing's relevance, which is already solid, enough to help the search upstart further nip at Google's search share?
And how will Matchbox fit into the Bing personalized search Microsoft introduced last month? When a user searches in Bing for the same search term on multiple occasions, and clicks on a link presented further down on the results page, the search engine will start bringing that link to the top of the page.
Google isn't standing still on this front. The company is building new search results based on contextual discovery, or leveraging users' locations and personal preferences to deliver custom results.