Community detection and semantic analysis on Twitter. The case of ‘no green pass’ and ‘no vax’ movement in Italy

The most extreme positions against anti-pandemic government measures are concentrated in the most democratic countries. Therefore, on the one hand, freedom of speech is guaranteed, being the essence of modern democracy; on the other stands the awareness of the risks that this freedom implies for the entire community. We are facing what some scholars call "the dilemma" of democracies. Increasing the drastic consequences of this dilemma is undoubtedly the communicative power of social networks. In this work, starting from data collected on Twitter from September 25, 2021 to October 22, 2021 (that is the period when local elections in many important cities of Italy took place) concerning the green pass debate in Italy, we construct a two-mode Semantic Network, which is a bipartite graph that describes connections between two types of nodes, social actors (in our case Twitter users) and semantic concepts. We analyze such network with detect communities of users and concepts, we use a proper two-mode community detection approach, i.e. an extension of the fast-greedy suited for the bipartite network called “DIRTLPAwb+”. The aim is to identify communities of users expressing different opinions and concepts within the green pass debate. Then, a Textual Correspondence Analysis will express the difference in lexicon usage with respect to the tweets belonging to the different communities. Results show that the combination of the two techniques can shed light on the differences or the proximity among groups of users posting on a given trending topic