EMOTIONAL TEXT MINING OF SOCIAL NETWORKS: THE FRENCH PRE-ELECTORAL SENTIMENT ON MIGRATION
F. Greco, D. Maschietti, A. Polli
Migration has actually gained considerable relevance both in the national and European political agendas and in general public debate. It is a challenge for governments, which needs coordinated responses to ensure citizens’ security. Furthermore, the terrorist attacks against western countries have called into question freedom of movement and residence for people within the European Union. In the electoral campaign, the populist rhetoric on migration has largely exploited citizens’ perception of insecurity, as in the French presidential electoral campaign of 2017. In order to analyse public sentiment on migration, we collected with twittweR of R Statistics a sample of 111767 messages containing the word “migrant” produced in the last two weeks before the first-round votes from the Twitter repository. The messages were collected in a large size corpus of over two million tokens to which we applied multivariate techniques, i.e. cluster analysis with a bisecting k- means algorithm and a correspondence analysis on the keyword per cluster matrix, in order to identify the contents and the sentiments behind the shared comments. The results show how the clusters and the factorial space are representative of the different ways of emotionally representing migration, highlighting the relevant aspects perceived by those who choose to express themselves through Twitter.