Hate Speech Detection using MLP and Random Forest

Social media platforms are essential parts of today’s communication because they help people of different countries to exchange information. However, due to the fact that these venues are unmonitored, hate speech has risen and this is a problem for both those to meet online and the general society. Because of the excessive involvement of social media in society, racisms both in old and new form have been observed. It is discovered that racism exists in several forms on social media, explicit and implicit. It can be observed by memes, and posting statements by racists with the intention of inciting hatred, violence, terrorism and fear in the society. Even though racial prejudice and discrimination are often associated with such expression still, there are indefinite and insulting statements on race, gender. Many countries regard sexual orientation as a threat to their political, social, and cultural stability, and to international security. The racist ideas propagate fast on social media, therefore social media must be closely monitored, and racist remarks must be promptly seen and set a ban on it. In this work, we present an approach involving deep learning and machine learning for classifying as well as detecting hate speech in the social media platforms. By utilizing a dataset of labelled tweets we preprocess the data, extract relevant characteristics, and train the model to categorize particular samples of hate speech, offensive language, and neutral content. We apply state-of-the-art deep learning methodology based on multilayer perceptron (MLP) neural networks, and machine learning technique which is random forest. We demonstrate that our algorithms are properly recognizing hate speech based on parameters such as accuracy and loss. This study supports the continuous efforts to prevent hate speech on social media and for general development of the safer and secure environment for the consumers.