No news , bad news ! A Personalized News Recommendation System

The development of the internet has drastically changed the newspaper industry, and the way people read news stories [3]. Newspapers are shifting from print to online version, with some unexpected results [1, 2, 6]. Good news for the newspaper industry? The proliferation of news aggregators such as Google News1 is blamed for killing the “old media”. With these aggregators, people visit a newspaper website to read one news story, then vanish off to another newspaper website to read another story [5]. As a result, the newspaper industry is dying because it is loosing parts of its advertising revenues, unable to retain its online users [6]. One possible response is to provide paid content, at the risk of jeopardizing its online advertising revenue [4]. But why would you pay for a news which is free on another newspaper website? Another solution would be to charge aggregators for a share of their advertising revenues. However, aggregators have no incentives to do so. News recommendations are one way to increase user retention. At the AI lab, we created a model for news recommendation and developed an efficient learning algorithm which recommends news to users based on their click behaviour. The goal of this project is to improve this model in order to generate better recommendations, and possibly propose a new approach.