Automatic Detection and Analysis of Extremist Video Content on Social Media Using Text

Abstract

Our prior work on radicalization focuses on Increasing our understanding of online radicalization efforts from analyzing right and left-leaning group videos in social media including groups from outside the USA. We studied what viewers found the videos appealing, what techniques were employed in producing radical content, and how we could computationally identify these techniques. More recently we have worked to identify potentially effective de-radicalizing techniques and worked to build a unified multimodal processing pipeline toolkit to help us create conversational systems on social media that are useful in de-radicalizing viewers.