UCLA

Online video content has proliferated as a key source for information in the form of video lectures, vlogs, reviews, how-to’s, and more. A video hosting service, YouTube, is now the second most popular search platform and reaches 81% of internet users under 20. However, most online videos do not provide audio descriptions (AD), presenting serious barriers for blind and visually impaired (BVI) people who may lack access to the visuals.In this thesis work, I present Human-AI systems to make online videos more accessible. Specifically, by (1) understanding what makes videos accessible to blind and visually impaired people; (2) automatically surfacing already accessible video content on platforms like YouTube with computer vision and natural language processing algorithms; and (3) helping video authors efficiently detect and address visual and auditory accessibility issues in their video authoring process.