Sunday, May 31, 2009

Challenges of customization of online news

Warning:- Technical post ahead.

When I came across this article sometime back, I was specially intrigued by one particular portion –

Everyone can publish, and everyone will - This is a problem, but since online journalism is still in its relative infancy it’s one that can be solved (we’re technology optimists, remember?). The experience of consuming news on the web today fails to take full advantage of the power of technology. It doesn’t understand what users want in order to give them what they need. When I go to a site like the New York Times or the San Jose Mercury, it should know what I am interested in and what has changed since my last visit. If I read the story on the US stimulus package only six hours ago, then just show me the updates the reporter has filed since then (and the most interesting responses from readers, bloggers, or other sources). If Thomas Friedman has filed a column since I last checked, tell me that on the front page. Beyond that, present to me a front page rich with interesting content selected by smart editors, customized based on my reading habits (tracked with my permission).

This got me thinking about the way presentation of news is expected to evolve in news websites, and how it would differ from offline newspapers. Most news websites these days do support “Guided navigation” features, suggesting articles to readers though links like “You might also be interested in..”, “People who read this also read..” etc. While “guided navigation” does have its obvious benefits, I have a difference of opinion with the suggestion that news websites of the future can completely replace the reading pleasure of offline newspapers. Anyone suggesting that content on a news website should ONLY be tailored to the interests and reading patterns of the reader, forgets that one of the key benefits of reading an offline newspaper is the ability to BROWSE through all topics all the time, and figure out what is interesting to you at that moment. For example, Mr. X may never have read the Finance section of the news website in the past, but if he sees an article about lay-offs or pay-cuts in the Finance section one day, he would definitely read it. Similarly, Mr. X might never read the Science and Health section, but might be interested in an article on an ailment which he suffers from. If the content of presented to Mr. X by the news website is tailored only on what Mr. X have been reading since the time he started using the service, such articles would never show up for him. Such algorithms would only narrow down a reader’s vision and knowledge acquisition to fields he or she has been 'historically' interested in (historically here means the history carried by the service provider), which might be biased by the previous set of articles and news items accessed on that service. Hence, an attempt of this kind, narrows the reader’s interest and intellectual growth in the direction defined by the interest he or she had from the time the reader started using the service. I do understand that such algorithms do get more “intelligent” as the duration of use increases. But there is a limit to which an algorithm can guess what a person wants to read.

A hybrid approach might be the way to go in the future. Front pages of news websites can have 2 sections, one presenting the latest headlines and editorial articles belonging to various topics (that is same for all readers), and one that presents articles based on historical data about the reader’s reading habits (stored with the reader’s permission). Going forward, algorithms of the future, apart from a person’s reading habits, must also consider other attributes like a person’s background (educational, professional, personal, demographic) while coming up with customized articles of the reader’s interest. This approach might help counter the two scenarios mentioned above. In this aspect, integration with social networking sites like Facebook, LinkedIn etc. might be an option worth considering.

No comments: