Hate Speech Detection Using Social Media Discourse: A Multilingual Approach with Large Language Model (African Journal of Biomedical Research)

Online social networks and microblogging websites have revolutionized communication, but have also seen an increase in online hate speech, which can provoke violence and societal division. Hate speech based on race, gender, or religion poses mental health risks and worsens social issues. Although current protocols have reduced overt hate speech, implicit forms have emerged, making detection challenging. This study focuses on detecting hate speech using social media discourse by creating a multilingual dataset in Urdu and English and applying various machine learning models, including GPT-3.5 Turbo. The study found that large language models are effective in detecting both explicit and implicit hate speech. The proposed methodology achieved a high accuracy of 0.91 and outperformed BERT by 5.81%. This research contributes to multilingual NLP and offers insights for reducing hate speech and promoting respectful communication.

https://www.researchgate.net/publication/388829147_Hate_Speech_Detection_Using_Social_Media_Discourse_A_Multilingual_Approach_with_Large_Language_Model