Microsoft’s Answer to PageRank: BrowseRank

Josh Catone

According to CNET a new paper out of Microsoft Research Asia (PDF) details what may eventually be Microsoft’s answer to the Google PageRank algorithm that was in a large part responsible for the Mountain View-based company’s ascension to the search engine throne. Microsoft’s version, called BrowseRank, would rank pages based on user behavior and not based on linking.

The basic idea behind Google’s PageRank is that the more a page is linked to, the most important it must be. Microsoft says that link analysis algorithms like PageRank are flawed, though, because they’re easy to be gamed and don’t take user behavior into consideration. Of course, Google’s actual implementation of PageRank is far from that simplistic and the company updates its search algorithms hundreds of time each year. Further, Google reminds us often that PageRank is just one of many things that it uses to rank search results.

Still, Microsoft thinks that it can do better — and it better hope that it can do a lot better. As we discussed earlier beating Google with technology means you have to beat the pants off of them and really wow users with dramatically better search results.

Microsoft Research Asia’s BrowseRank algorithm ditches the link graph model that was popularized by Google, and instead creates a user browsing graph that looks at things like which links users clicked on and how long they stayed on each page.

“User behavior data can be recorded by Internet browsers at web clients and collected at a web server,” according to the researchers. Microsoft Research Asia said they gathered anonymous data from an “extremely large group of users under legal agreements with them” to put their theory to the test. The idea is that you can take anonymous browsing data from hundreds of millions of users and create a user browsing graph that can paint a picture of which pages are most important to users.

“The user browsing graph can more precisely represent the web surfer’s random walk process, and thus is more useful for calculating page importance. The more visits of the page made by the users [sic] and the longer time periods spent by the users on the page, the more likely the page is important,” say the researchers. “With this graph, we can leverage hundreds of millions of users’ implicit voting on page importance. In this regard, our approach is in accordance with the concept of Web 2.0.”

Of course, by itself, user browsing behavior probably isn’t enough to rank pages — if BrowseRank was used on its own, it would be easy to see MySpace and Facebook and video sites like Hulu shoot to the top of search results pages. However, Microsoft researchers think that it could be combined with other web page ranking algorithms to greatly enhance search results. “It is also possible to combine link graph and user behavior data to compute page importance,” they write. Researchers said that initial results from their tests using BrowseRank showed better performance than existing methods.

It wouldn’t be surprising to learn that Google had something similar under development. Google is already capturing user browsing behavior via its popular Google Toolbar, and appears to have put some of that data to use earlier this year with the launch of Ad Planner and enhancements to Google Trends that include web traffic. Using that data in search engine results rankings — or at least experimenting with doing so — isn’t a huge leap. Google is hardly a sleeping giant.