We investigate the complex relations existing within news content in the 27 countries of the European Union (EU). In particular we are interested in detecting and modelling any biases in the patterns of content that appear in news outlets of different countries. We make use of a large scale infrastructure to gather, translate and analyse data from the most representative news outlets of each country in the EU. In order to model the relations found in this data, we extract from it different networks expressing relations between countries: one based on similarities in the choice of news stories, the other based on the amount of attention paid by one country to another. We develop methods to test the signiﬁcance of the patterns we detect, and to explain them in terms of other networks we created based on trade, geographic proximity and Eurovision voting patterns. We show that media content networks are 1) stable over time, and hence well deﬁned as patterns of the news media sphere; 2) signiﬁcantly related to trade, geography and Eurovision voting patterns; 3) by combining all the relevant side information, it is possible to predict the structure of the media content network. In order to achieve the above results, we develop various pattern analysis methods to quantify and test the non-metric, non-symmetric pairwise relations involved in this data. These methods are general and likely to be useful in many other domains.
|Translated title of the contribution||Predicting relations in news-media content among EU countries|
|Title of host publication||Cognitive Information Processing (CIP), 2nd International Workshop on|
|Publisher||Institute of Electrical and Electronics Engineers (IEEE)|
|Publication status||Published - 2010|
Bibliographical noteOther page information: 269-274
Conference Proceedings/Title of Journal: Cognitive Information Processing (CIP), 2nd International Workshop on
Other identifier: 2001269