Monitoraggio della fiducia sui vaccini analizzando un gruppo di discussione online

The initial idea of this thesis work is based on a phenomenon called vaccine hesitancy. Vaccine hesitancy is the reluctance to vaccinate despite the availability of vaccines. This happens mostly in industrialized country like USA, several states in Europe where vaccines are available and sometimes free for most of them (e.g. Italy). The thesis analyses a vaccines discussion on a web site forum to determine contents of treated topics among users. For this purpose we chose a web forum where the communication among users is as much as possible similar to a real discussion among people, unlike in Twitter or Facebook where the opinion of users are more polarized. These two platforms allow the emergence of the echo chamber on vaccine debate, clustered and enclosed communities where the exposure of opposing views is very limited. Far away from this phenomenon, choosing suitable forum, we try to understand how a user tends to communicate with others according to the discussed topics on the interactions network, reconstructed through the exchange comment-post. After a brief introduction of the vaccines origin and an explanation of the hesitancy phenomenon, the thesis is composed by 3 parts: i) description of the theoretical framework used to analyse data; ii) the research methods and iii) original results from the analysis. The first part consists in a brief introduction to network theory and theoretical framework of the applied machine learning techniques (Latent Dirichlet Allocation and Doc2Vec). The second one describes: the methods (web scraping) used to collect public data (posts and comments about vaccines), the conducted analysis and the structural characteristics of the reconstructed network. In the third part instead we show the temporal distribution of our data, the found threads about vaccines showing the possible lack of echo chamber and likely a more discussion among opposing views. We found a significant correlation between the users interactions and the topic they discuss.