A Deep Learning Approach of Hate Speech and Offensive Language Detection on Twitter -A REVIEW

computer engineering

Recent years have seen an increase in the prevalence of hate speech, abusive language, misogyny, racism, cyberbullying, and other forms of abuse on Facebook, Twitter, and other social media platforms. People are more likely to propagate this type of action to disparage or damage someone's reputation. Such violent and offensive behaviour has grown enormously, as evidenced by. These occur as a result of people's freedom or openness to express themselves on social media platforms without fear or regard for the feelings of others. These platforms lack the capacity to effectively address the issue of online abuse, hate speech, and offensive language on their platform. Many other companies, research organizations are investing lots of money and research effort to curb this problem but they don’t get much success because there is a need of great manual work to detect and remove online posts having hate speech or offensive language. The main challenge for automatic detection of hate speech on social media is to distinguish it from offensive language, cyberbullying and another form of abuses. In our research, we introduce deep learning techniques to identify hate speech and objectionable language on Twitter. These techniques include CNN with global and average max pooling, CNN with dynamic convolution neural networks with k-max pooling, and multi-layer perceptrons. We tested these models experimentally using four publicly available Twitter hate and abusive datasets (largest twitter dataset till). On three of the four datasets, our model DCNN with k-max pooling and MLP produced state-of-the-art results. In general, our models performed better on these datasets and produced an excellent outcome when compared to earlier research on the same dataset.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.