What is it about?
This research focuses on automatically detecting hateful content in internet memes and videos by analysing both visual elements (like images or video clips) and textual information (like captions or speech). We explore different ways of combining this multimodal data to identify offensive content better, helping platforms manage harmful material online effectively.
Featured Image
Photo by Jon Tyson on Unsplash
Why is it important?
Online hate speech, especially in formats like memes and videos, has grown significantly and poses serious societal challenges. Our work uniquely addresses this by effectively combining visual and textual information through advanced multimodal techniques. This approach allows for more accurate detection of harmful content, helping platforms better protect users from offensive and hateful material in a timely and efficient manner.
Perspectives
Working on this research has been deeply motivating for me, as addressing hate speech aligns closely with my commitment to fostering safer and more inclusive digital spaces. Through this project, I have come to appreciate the complexity of interpreting multimodal content and how powerful AI techniques can significantly impact real-world issues like online harassment and hate speech. I am optimistic about the practical applications and positive changes this technology can support.
Girish Koushik
University of Surrey
Read the Original
This page is a summary of: Towards a Robust Framework for Multimodal Hate Detection: A Study on Video
vs.
Image-based Content, May 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3701716.3718382.
You can read the full text:
Resources
Contributors
The following have contributed to this page
