DejaVu: A System for Journalists to Collaboratively Address Visual Misinformation

Journalistic work increasingly depends on information from web sources and social media. Visual misinformation, for example images that have been manipulated or taken out of context, pose a signiicant issue for journalists using these sources. Based on informal interviews with working journalists , we developed DejaVu, a system that supports journalists in the task of detecting visual misinformation. De-jaVu streamlines the task of looking for near-identical image matches using reverse image search, and extends it by crawling and indexing rogue social media sites such as 4chan. More importantly, DejaVu supports collaboration between journalists by allowing them to oag images, which are then indexed such that the image and its near-duplicates are highlighted for other journalists regardless of where on the Web they ynd them. A preliminary evaluation of DejaVu's visual indexing shows that it can support such collaboration even when nagged images are re-posted aaer being further manipulated.