Abstract: Hostile discourse, characterized by discriminatory language, expressions of hate, or overt aggression based on individual or group identity, presents a formidable challenge in online communication. This article is an in-depth study of hate speech research, specifically the definition and classification of hate speech in text. Through a comprehensive review, the research explores various techniques, ranging from classical machine learning algorithms to advanced deep learning models such as convolutional neural networks, short-term memory networks, gated recurrent units, and transformer-based architectures, with a special focus on bidirectional LSTM with self-generating mechanisms and feedforward neural networks. Moreover, the paper offers practical insights for effective model development, emphasizing the necessity of harnessing large-scale social media datasets, ensuring data balance for representative training, implementing regularization techniques for improved generalization, and incorporating a validation set for accurate performance evaluation. By combining theories from a variety of research methods and using an integrated approach from diverse models, this study aims to provide researchers and practitioners with a conceptual framework for developing powerful models that are effective. In summary, this article highlights the importance of adapting technology to the dynamic field of online communication, with the overall goal of promoting security and benefiting diverse communities..
Keywords: Hostile Discourse Detection, Hate Speech, Machine Learning Models, Data Preprocessing, Model Training, Ensemble Approach
|
DOI:
10.17148/IJARCCE.2024.13558
