What Readers Want




CAPTION:The word cloud shows in pink the words that increase the appeal of a text for readers of the website “News.com.au”. The size of a word in the image reflects the magnitude of this effect. Notice the references to people and to the location of the outlet in Australia.


A new study has analysed the millions of choices available to readers of online news and created a model to find out ‘what makes people click’. The researchers developed a model of “news appeal” based on the words contained in an article’s title and text intro, which is what a reader uses to choose to click on a story. The study by academics at the University of Bristol’s Intelligent Systems Laboratory is published in Pattern recognition - applications and methods.

The aim of the study was to model the reading preferences of the audiences of 14 online news outlets using machine learning techniques. The models, describing the appeal of a given article to each audience, were developed by linear functions of word frequencies. The model compared articles that became “most popular” on a given day in a given outlet with articles that did not.

The research, led by Nello Cristianini, Professor of Artificial Intelligence, identified the most attractive keywords, as well as the least attractive ones, and explained the choices readers made.

The team created a model for each user group they had data on, including online news for the BBC, Forbes and Australian newspapers. After scoring articles by reader preferences, the researchers then ranked the articles by their appeal, and studied what might explain the choices online readers make.

Professor Cristianini, speaking about the research, said: “We found significant inverse correlations between the appeal to users and the amount of attention devoted to public affairs. People are put off by public affairs and attracted by entertainment, crime, and other non-public affairs topics.”

The researchers examined over two million articles, collected over nearly eighteen months. The models analysed user choices, choices were then used to compare both the audiences and the contents of various news outlets. The researchers found that there is a significant correlation between the demographic profiles of audiences and their preferences. They also found that content appeal is related both to writing style - with more sentimentally charged language being preferred, and to content with “public affairs” topics, such as “finance” and “politics”, being less preferred.




CAPTION:The most popular words are colored in pink, and the least popular in black. Based on Forbes.


VIDEO: This video is (also) about the present study: http://www.see-a-pattern.org/?q=content/patterns-media-content

IMAGES: High Resolution Images

List of publications:

  • Modelling and Explaining Online News Preferences.
    Elena Hensinger, Ilias Flaounas, Nello Cristianini.
    In Collection: Pattern Recognition - Applications and Methods, Springer Berlin Heidelberg, 2013. http://dx.doi.org/10.1007/978-3-642-36530-0_6 [PDF]
  • The Appeal of Politics on Online Readers. Elena Hensinger, Ilias Flaounas and Nello Cristianini, Presented at "Internet, Politics, Policy 2012: Big Data, Big Challenges?", Oxford Internet Institute, September 2012. http://microsites.oii.ox.ac.uk/ipp2012/papers
  • Modelling and Predicting News Popularity. Elena Hensinger, Ilias Flaounas and Nello Cristianini. Pattern Analysis and Applications, Springer, 2012.
  • What makes us click? Modelling and Predicting the Appeal of News Articles. Elena Hensinger, Ilias Flaounas and Nello Cristianini. In Proceedings of the 1st International Conference on Pattern Recognition Applications and Methods, 2012.
  • Learning Readers’ News Preferences with Support Vector Machines. Elena Hensinger, Ilias Flaounas and Nello Cristianini. In Proceedings of Adaptive and Natural Computing Algorithms - 10th International Conference, ICANNGA 2011.
  • Learning the Preferences of News Readers with SVM and Lasso Ranking. Elena Hensinger, Ilias Flaounas and Nello Cristianini. Artificial Intelligence Applications and Innovations (AIAI), Springer Boston, Cyprus, 2010.