<![CDATA[ A The rapid growth of social media has increased the risk of hate speech proliferation, driving extensive research in Natural Language Processing (NLP) to develop more accurate automatic detection systems. Over the past decade, hate speech detection approaches have evolved significantly, shifting from traditional machine learning methods to deep learning architectures and advanced transformer-based models. However, comprehensive bibliometric studies that map methodological developments and implementation domains remain limited. This research analyzes 1,335 publications indexed in Scopus to identify trends in methodological approaches (e.g., SVM, Naive Bayes, LSTM, and BERT-family models) and application domains (Twitter, Facebook, YouTube, and multilingual contexts). Keyword extraction and temporal trend visualization were conducted using Python. The findings indicate that transformer models have dominated research since 2020, accompanied by a shift from single-text analysis toward multimodal and multilingual approaches. This study highlights future research directions involving transformer integration, multilingual processing, and Explainable AI (XAI) to enhance transparency in hate speech detection.]]>
Copyrights © 2026
