Journalists in newsrooms across the world depend on images sourced from the web and social media when reporting stories. But images can be altered, misrepresented, or taken out of context, making discovering this "visual misinformation" a vital task for journalists. The problem is complicated by the necessity of identifying the first appearance of an image, searching un-indexed social media sites like Reddit and 4chan, and sharing information between journalists.
Cx faculty members Mor Naaman (Cornell Tech) and Ofra Amir (Technion), along with researchers Hana Matatov, Adina Bechhofer, and Lora Aroyo have been tackling this issue. The team interviewed New York City-based journalists and information workers about their content verification needs. Based on their findings, they built a system called DejaVu, designed to support journalists in searching for, detecting, and annotating visual misinformation.