A Deep Learning Approach for Hate Speech Detection on Twitter: Comparative Analysis with Traditional Models

Abstract

In the modern era of technology, social media platforms have become the primary channels for communication between people and organizations and for sharing and disseminating information, but the open nature of social media platforms makes them a fertile ground for harmful content, including hate speech. Hate speech is a form of expression that incites or encourages discrimination, hatred, or violence against individuals or groups based on certain characteristics, including but not limited to race, religion, gender, or sexual orientation, and has become a growing concern, making the spread of such content pose significant challenges to online communities and social cohesion. In this study, we propose a deep learning-based model to classify Twitter tweets into three categories: hate speech, neither, and abusive. The proposed model relies on the use of a multi-layer neural network with a set of dropout layers to enhance the generalization of the model and avoid the problem of overfitting. The performance of the proposed approach was analyzed and contrasted with traditional machine learning (ML) algorithms, including decision tree, random forest, and support vector machine (SVM). The experimental results showed that the proposed model reached the best accuracy of 97%, significantly outperforming the basic machine
learning algorithms. The results confirm the effectiveness of the proposed model and its ability to capture the underlying patterns and semantic and linguistic complexities within the tweet text content, which enables it to classify hate speech with high accuracy.