AUTOMATIC DETECTION AND ANALYSIS OF EXTREMIST VIDEO CONTENT ON SOCIAL MEDIA USING TEXT, SPEECH, AND VISUAL FEATURES
Abstract
Radicalization is the process of developing extremist ideologies and beliefs in others (Borum, 2011). In recent years, such efforts are increasingly seen on social media, where extremists spread their ideology and attempt to influence others to share extremist beliefs about race, ethnicity, gender, or religion through text-based posts, images, and videos. Previous work has proposed many theories of how and why radicalization develops, but little has been done to empirically test these theories on a large scale and to answer questions about specific features of radical messages that are significantly correlated with success in attracting followers. Some simple statistical efforts have employed word lists, e.g. in Twitter messages or Reddit posts, to use in collecting and assessing radicalizing data; however much of current radicalization is being attempted on sites like YouTube, through shared videos, or on groups’ own websites, where many features beyond words can be used to attract viewers, including audio, speech, and visual data. These additional features (e.g. visual memes) have been shown in qualitative analyses to play a large role in radicalization so they should be very useful additions to the feature-set.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 12, 2021
- Source ID
- FA95502010400
Entities
People
- Julia Hirschberg
Organizations
- Air Force Office of Scientific Research
- Trustees of Columbia University in the City of New York
- United States Air Force