Bangla Hate Speech Detection in Videos Using Machine Learning

Abstract

Due to the internet and social media advance, people now communicate their ideas and opinions on many topics freely through several channels. However, because of its ethnicity, religion, genders, etc., this freedom of expression is exploited to direct hate to individuals or groups of people. The increase in hate speech led to disputes and cyberbullying, which drove many companies to find ideal solutions to this problem. The strategy we have taken is to use machine learning classification methods to classify videos. We have generated our dataset and tested it with several training models in this paper, utilizing the spoken material of the videos from YouTube. We experimented with some of the machine learning models along with some deep learning techniques. Our experiments concluded that the logistic regression and deep learning GRU model bring out the best accuracy on the current dataset.