Hate Speech Detection using Global Vector and Deep Belief Network Algorithm

Hate speeches are words, behaviors, and actions prohibited because they lead to acts that trigger violence and anarchist attitudes towards other individuals or groups. Since the 2014 presidential election, the term 'hater' has been widely known, marking people with a tendency to practice speech utterances in certain people and groups. Thus, the internet's ethics need to be emphasized, considering that the internet is a necessity for today's society. Nevertheless, more and more users are also many parties who abuse the internet to spread information about speech utterances such as ethnicity, race, and religion. In this project, a system will be created to detect hate speech in the form of tweets on twitter. The author's method is the Deep Belief Network method by weighing the Global Vector feature to increase accuracy before classification. The making of this system is expected to be able to find out and detect hate speech from the text previously in the form of tweets. By using the Deep Belief Network method, the results of this study were obtained with an accuracy =86,00%, precision =82,00%, recall =89,13% and Fl-Score =85,42%. After doing this research, it is expected that the computer can find out and classify the existence of hate speech in the text.