Sexism Identification on TikTok: A Multimodal AI Approach with Text, Audio, and Video

Sexism persists as a pervasive issue in society, particularly evident on social media platforms like TikTok. This phenomenon encompasses a spectrum of expressions, ranging from subtle biases to explicit misogyny, posing unique challenges for detection and analysis. While previous research has predominantly focused on textual analysis, the dynamic nature of TikTok demands a more comprehensive approach. This study leverages advancements in Artificial Intelligence (AI), specifically multimodal deep learning, to establish a robust framework for identifying and interpreting sexism on TikTok. We compiled the first dataset of TikTok videos tailored for analyzing sexism in both English and Spanish. This dataset serves as an initial benchmark for comparing models or for future investigations in this area. By integrating text, linguistic features, emotions, audio, and video features, this study identifies unique indicators of sexist content. Multimodal analysis surpasses text-only methods, particularly in understanding the intentions behind sexism.