Abstract

In this report we describe the results of the analysis that we performed through topic modeling on the texts that we collected and described in the previous deliverable 5.2. In particular, our aim was to analyze the latent meaning structure, and the shared meanings of EU policies and EU identity, on four levels of communication: 1) Communications of the EU: here we analyzed the magazine Panorama; 2) Local implementation: here we analyzed financed projects’ abstracts; 3) Local press: at this level we focus on newspapers; 4) Social media: at this level we focus on Facebook. The analysis performed with Topic Modeling on the corpus of tweets collected from Twitter did not provide good results, due to the short length of each tweet. It is a result that is expected, but collected data are not useless. In order to provide a better coherence of this report, which focuses on the results of Topic Modeling, we are therefore not providing an analysis of tweets here. On the contrary, we will analyze tweet in the next deliverable 3.3, that is aimed at analyzing the use LMAs make of social media. With regards to the other level, we are presenting 16 models: one model for Panorama, one model for financed projects’ abstracts, 7 models for newspapers – one for each country – and 7 models for Facebook. Regarding Facebook, LMAs in our case study region in UK do not have a Facebook profile. The seventh model regards thus European Institutions’ Facebook profiles. Each model is composed of a list of 20 topics, which we analyze and characterize through descriptive statististics. We focus in particular on the emergent topics related to European Identity and Cohesion Policy. Moreover, in this report, we make use of formal methods and techniques to visually represent the meaning of topics elicited through topic modeling. The following paragraph provides a short introduction to the technique of topic modeling. Then four paragraph accounts for the several topic models elicited at different level of communication. Finally, conclusions make sense of the whole analysis.