“This paper describes the detection and classification of hate speech on Twitter by using deep learning. This can fill in the gaps between current models with higher accuracy and reliability. Thus, this research is beneficial in several ways, such as the detection of hate speech in distinct categories, such as in toxic, severe toxic, obscene, threat, insult and identity hate. The developed application used deep-learning algorithms to find the number of occurrences of the words and semantic words. The LSTM model is used to train the data set and to get the probability values. The classes of hate speech were calculated against the training data set and were found to be above 72%. In conclusion, the developed method can help to detect and classify hate speech into six classes on Twitter.”
http://ir.kdu.ac.lk/bitstream/handle/345/5186/1.pdf?sequence=1&isAllowed=y