Code for Caption Crowd (IKILeUS)

CaptionCrowd is an interactive platform developed within the IKILeUS project at the University of Stuttgart to improve video caption accuracy for the Deaf and Hard of Hearing (DHH) community. While automatic captions provide some accessibility, they often contain errors in grammar, homophones, and domain-specific terminology, making comprehension challenging. CaptionCrowd enables users to collaboratively identify and correct inaccurate captions in real-time, improving their quality through community-driven feedback.<br>
The platform features a user-friendly web-based video player that allows users to highlight incorrect words in subtitles, with their selections recorded for further analysis. User testing with 16 participants revealed that manually correcting captions can be cognitively demanding, highlighting the ongoing need for enhanced accessibility solutions.

CaptionCrowd是由斯图加特大学IKILeUS项目开发的交互式平台，旨在面向聋人与重听（Deaf and Hard of Hearing, DHH）群体提升视频字幕准确率。自动字幕虽能提供基础的无障碍支持，但常存在语法、同音异义词以及专业领域术语错误，导致理解难度上升。CaptionCrowd支持用户实时协作识别并修正不准确的字幕，通过社区驱动的反馈机制提升字幕质量。
该平台搭载界面友好的网页端视频播放器，允许用户标记字幕中的错误词汇，用户的选择记录将被留存以用于后续分析。针对16名参与者开展的用户测试显示，手动修正字幕会带来较高的认知负荷，这也凸显了对更完善的无障碍解决方案的持续需求。