Multidimensional and Multilingual Emotional Analysis

In order to monitor informal political online discussions and to lead a better understanding of hate speech on social media, we found that it was necessary to use sentiment quantification for languages with few training datasets. Previous studies mainly rely on languages with enough data to train a model. Several statistical and machine learning models were produced and compared in three languages (English, Portuguese and Polish). This work shows promising results when inferring sentimental dimensions, even in languages other than English.