Verbal Aggressions Detection in Mexican Tweets

Abstract

Verbal aggressions are a struggle that a great number of social media users have to face daily. Some users take advantage of the anonymity that social media give them and offend a person, a group of people, or a concept. The majority of proposals which pretend to detect aggressive comments on social media handle it as a classification problem. Although there are a lot of techniques to face this problem in English, there is a lack of proposals in Spanish. In this work, we propose using several Spanish lexicons which have a collection of words that have been weighted according to different criteria like affective, dimensional, and emotional values. In addition to them, estructural values, word-embeddings and one-hot codification were taken into account.