A diversity of techniques has been used for monitoring public opinion, allowing to analyse the evolution of narratives (chains of concepts and arguments) being propagated in society. In this session, a new quantitative method for discourse analysis will be presented, that is well adapted for large textual corpus. This method is based on network theory, machine learning, natural language processing, and mathematical modelling of multi-level complex systems. The method is applied to the largest European newspapers during 2016. It allows characterising the evolution of media coverage in different European countries regarding migration and humanitarian crisis. Namely, narratives propagated by newspapers in the South, North and East of Europe will be compared. Discussion will focus on the analysis of these results.

--