Towards the Detection of Hateful Sentiment in Social Networks

Hate speech in social networks is affected and therefore increased thanks to anonymity. The use of opinion mining and Natural Language Processing (NLP) have increased their activity during the last few years because they provide an approach to analyze people's opinion, attitude ratings and emotions in the evolution of web 2.0. This article shows the analysis of hateful sentiment in social networks, specifically on Twitter. For this purpose, the tweets have been obtained through the source of information provided by the Tweepy API, thus forming a corpus with tweets that will be labeled as hate and non-hate, as the input of the analysis. To carry it out, a series of tasks are performed: preprocessing, feature extraction, vectorization, training of the Naive Bayes classification algorithm and validation of the algorithm along several metrics. We conclude the validity of the method, which could be used to make a more precise specification of hate speech, with the aim of identifying social biases, such as gender or racial discrimination, among others.