Classifying possible hate speech from text with deep learning and ensemble on embedding method

Abstract

Hate speech can be defined as the use of language to express hatred towards another party. Twitter is one of the most widely used social media platforms in the community. In addition to submitting user-generated content, other users can provide feedback through comments. There are several users who intentionally or unintentionally provide negative comments. Even though there are regulations regarding the prohibition of hate speech, there are still those who make negative comments. Using the deep learning method with the long short-term memory (LSTM) model, a classifier of possible hate speech from messages on Twitter is carried out. With the ensemble method, term frequency times inverse document frequency (TF-IDF) and global vector (GloVe) get 86% accuracy, better than the stand-alone word to vector (Word2Vec) method, which only gets 80%. From these results, it can be concluded that the ensemble method can improve accuracy compared to only using the stand-alone method. Ensemble methods can also improve the performance of deep learning systems and produce better results than using only one method.