Revealing Political Sentiment with Twitter: The Case Study of the 2016 Italian Constitutional Referendum
Marco Campanale, Enrico Giacinto Caldarola · 2018
In this work we argue about the use of Twitter as a Palantir for predicting and revealing the political orientation of users in election campaigns. Specifically, our study aims at revealing the political orientation of a Twitter user in the context of the 2016 Italian Constitutional Referendum. After having collected and processed over 1,200,000 tweets, we classified them as YES-oriented, NO-oriented or UNCERTAIN, by exploiting the Naive Bayes Multinomial Text classification algorithm. We found that Twitter is used massively for political deliberation and just by counting the messages related to a political party we can reveal the election result. Moreover, the words used by the YES party and the NO party twits are in line with the language used by politicians in the real word and this strongly supports our analysis. Furthermore, our approach can be applied to different scenarios, in which it is necessary to distinguish between two main classes and another UNCERTAIN class. Finally, the results obtained by our classification methodology are very promising and encourage us to continue investigating this topic, deriving suggestions for further research.