Comparative Analysis of Machine-Learning and Deep Learning Algorithms Using Manta Ray Foraging Optimization for the Detection of Hate Speech

User safety and societal well-being are two key concerns given the rising hate speech on online platforms. This study aims to tackle hate speech detection (HSD) using the Davidson’s dataset; a popular benchmark in this field. Count vectorization is used to preprocess textual data thereby enabling extraction of meaningful features. Various machine learning (ML) and deep learning (DL) architectures such as Support Vector Classifier (SVC), Long Short-Term Memory (LSTM), Bidirectional LSTM (Bi-LSTM), and Attention-based LSTM models are tested for performance. In addition, in this study a novel metaheuristic swarm-based optimization algorithm, Manta Ray Foraging Optimization (MRFO), has been used for enhanced feature selection to increase the detection accuracy of hate speech. The findings of the study have shown that MRFO in conjunction with attention-based approach produces an impressive performance.