Changing the language also changes the sentiment algorithm? Exploring the communication on Green Pass

The development of techniques and tools for automated sentiment analysis of digital content is an area that is receiving more and more attention from the scientific community (Punziano, 2021). For example, many researchers, especially scholars of web communication, have to deal with the experimentation of innovative methods that combine the need to operate on new data with reliable procedures that do not lose sight of the meaning (and sentiment) of the content. We know that the uncritical application of automatic analysis techniques can be insidious (Amaturo and Aragona, 2016; Kitchin 2014) for this reason the scientific community is very interested in the supervised approach: it is based on the probability that a content in a defined context expresses a specific feeling. These algorithms learn from the researcher's knowledge background and is a significant step to limit the risks of a data-driven approach with big data (Aragona, 2017; Kitchin, 2014). In a previous work (Acampa, De Falco, Trezza, 2020) we compared the three machine learning approaches to measure feelings in a corpus of tweets in Italian, the results obtained prompted us to test the approaches also on other languages (English, French, Spanish) with a view to comparison of results. The case study is the analysis of the sentiment expressed by Italians on Twitter regarding the adoption of the vaccination green pass. Considering the conflicting reactions of European citizens regarding this measure, we have assumed that the treatment of the issue is characterized by antithetical perceptions and feelings useful for constructing an optimal test of the models. To start the supervised model, you will need to build a layered sample of tweets and manually classify them. Finally, an index will be built on the tweets processed with the aim of detecting the goodness of the results produced and comparing them with all the languages tested. The comparison of the three algorithms helps us to understand not only what is the strategy to verify the quality of the data obtained.