Title: Time sensitive topic-based communities: The case of the vaccination debate in Italy
Authors: Rebecca Graziani - Bocconi University (Italy) [presenting]
Amelia Compagni - Bocconi University (Italy)
Abstract: The aim is to reconstruct the debate that was developed in Italy around the compulsory vaccination of children of school age and culminated in 2017 with the emanation of a decree law by the Italian government. We analysed all public statements released about the issue to ANSA, the most important news agency in Italy. The corpus was assembled by retrieving all ANSA statements between January 1st 2015 and January 21st 2019, using the word vaccini (vaccines). We created a corpus of 3,225 statements, that we annotated and manipulated so to analyse sub-corpora based on the author, organization or date and as such to compare the statements produced by different subsets of actors or at different time periods. With the aim of identifying the main topics in the debate, we ran a topic analysis, with the Latent Dirichlet Allocation approach. The solution with fifteen topics ended out to be the best one in terms of coherence measures and interpretability. We restricted our attention to a selection of authors and implemented a network analysis on the time series of estimated topic weights by authors, so to identify time-sensitive topic-based communities. The analysis leads to the identification of three thematic communities.