Advancing Hate Speech Detection: A Multilingual System Using LLAMA-2 for Real-Time Analysis of Audio, Video, and Text Content

This paper introduces an advanced, multilingual hate speech detection system designed to process various media formats, including audio, video, and text. Unlike traditional detection models that focus solely on text analysis, our system leverages OpenAI's Whisper model to convert audio and video content into text, enabling a broader scope of analysis. At its core, the system utilizes a fine-tuned Llama-2 model specifically trained for hate speech classification. The detection system is accessible through a user-friendly Streamlit web application, providing real-time processing and feedback. Results demonstrate that this comprehensive, multimedia approach significantly enhances detection accuracy and usability, marking a notable advancement in the field.