Paper Reading #19 - Personalized News

Comments:
Comment 1
Comment 2

References:
Title: Personalized News Recommendation Based on Click Behavior
Authors: Jiahui Liu, Peter Dolan, Elin Ronby Pedersen
Venue: IUI 2010, Feb. 7-10 2010

Summary:
In this paper, the authors describe a new method of recommending news articles to read on the Google News service. They do this by analyzing both the articles users choose to click on, as well as current news trends.

They begin by doing an in-depth analysis of the influences on what news readers are interested in. They then find that readers are not only influenced by their own personal tastes due to age, gender and job, but also that they are influenced by current news stories. Furthermore, the news stories that people have an interest in have a large correlation based on location.

From this data, they design two probabilistic models that show the user's likelihood of clicking a particular link. One model is based upon their individual clicks, which should correspond to their personal interests, and the other is based on news stories clicked upon by people who are nearby, to get an idea of current breaking news in the area. They then find that this new method caused a noticeable increase in clicks on Google News recommendations.

Discussion:
I thought this article was interesting mainly because it gave a sort of window into how they use user information at Google. I was impressed by how creative they are in using not only personal clicks but also those of the collective to get a larger picture of the data.

As far as the product goes, I think it's a positive move, since I definitely don't have a lot of time to spend searching through articles for what I'm looking for. I also wonder if they use a similar algorithm to this in Google Reader, since that is my main news hub.

(Image courtesy of: Google News)