MetaHate: Text-based Hate Speech Detection for Metaverse Applications using Deep Learning
Judith Nkechinyere Njoku, Anthony Uchenna Eneh, Cosmas Ifeanyi Nwakanma, Jae‐Min Lee, Dong‐Seong Kim · 2023
The rise of digital communication and the metaverse has revolutionized interaction paradigms, while also introducing challenges in ensuring safe engagements. Hate speech, pervasive in digital spaces, threatens inclusivity. This research introduces a tailored hate speech detection system for the metaverse. Through a comprehensive evaluation of deep learning models, effective real-time detection approaches are identified. To guarantee reliable deployment in the metaverse, models are subjected to explainability assessments using Local Interpretable Model-Agnostic Explanations (LIME). A lightweight Convolutional Neural Network (CNN) model is developed and deployed on the Roblox server, exhibiting commendable accuracy and efficiency. Remarkably, quantization reduces the CNN model size by 93.59%. This pioneering study addresses the dearth of metaverse-focused hate speech research, fostering secure and inclusive virtual spaces.