Paper Reading #19: Personalized News Recommendation Based on Click Behavior

Comments: Chris Kam, Derek Landini.
Reference Information:
Title: Personalized News Recommendation Based on Click Behavior
Authors: Jiahui Liu, Peter Dolan, Elin Rønby Pedersen.
Venue: IUI’10, February 7–10, 2010, Hong Kong, China.

Summary: In their research paper, the designers lament the current state of online news. Their aim is to develop a system in which a website can predict what type of news stories a user might want to read, without having to rely on the user to proactively provide their preference of what type of news they like to read.

Their system tracks a person around a news site, tracking their clicks and examining how long they spend on each page. It's designed to be mindful of user's changing tastes, and so it changes over time as the users give the system more information. In order to make sure that the results from the system are accurate, it had a polling mechanism in place to see if a user's interest in the customized article selection improves from the previously shown selection of stories.

After designing the system to take advantage of the Google News website, the researchers presented the software to a group of test users, and compared their results against a control group. After careful testing and measuring, they found that the system they had created consistently more relevant news to the user.

Discussion: This sounds like a great system, and I'm sure I'd love it if it was implemented on a news website I visited. However, the idea of a site so closely monitoring my behavior always rubs me the wrong way. It seems like there should be better ways of gathering information than watching a user's every move.