Detecting Macro-Patterns in the European Mediasphere

The analysis of the contents of news outlets has been the focus of social scientists for a long time. However, content analysis is often performed on hand-coded documents, which limits the size of the data accessible to the investigation and consequently limits the possibility of detecting macro-trends. The use of text categorisation, clustering and statistical machine translation (SMT) enables us to operate automatically on vast amounts of news items, and consequently to analyse patterns in the content of outlets in different languages, over long time periods. We report on experiments involving hundreds of European media in 22 different languages, demonstrating how it is possible to detect similarities and differences between outlets, and between countries, based on the contents of their articles.