Developing and Deploying a Real-World Solution for Accessible Slide Reading and Authoring for Blind Users
Zhuohao Zhang, Gene S-H Kim, Jacob O. Wobbrock · 2023 · Proceedings of the 25th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS '23) · doi:10.1145/3597638.3608418
Summary
This paper presents A11yBoard for Google Slides, a multi-device multimodal system that makes presentation slide reading and authoring accessible to blind users. Presentation software like PowerPoint and Google Slides relies on 2-D "artboard" canvases where objects are placed in arbitrary spatial arrangements with no inherent reading order—a paradigm fundamentally incompatible with screen readers, which are designed for linear content. The system consists of a Chrome browser extension that connects to Google Slides and an iOS mobile app that mirrors slides onto a touch screen device. Blind users can explore slides using a "reading finger" on the touch screen, receiving audio tones indicating object boundaries and Z-order layering, along with speech feedback describing object properties. The system supports three input modalities: touch and gesture for spatial exploration, speech commands for interpreting and generating content, and keyboard search for complex editing operations. The tool was developed through a participatory design process with a blind co-author (GK) over multiple iterative sessions, addressing six key design challenges including limited API control, multi-device screen reader interference, balancing efficiency with expressiveness, and supporting individual differences in perception. Two real-world field deployments with blind participants validated the system over five and seven days respectively.
Key findings
Both field deployment participants successfully read and recreated slide decks independently—tasks they described as previously "impossible" without sighted assistance. P1, who used presentation software weekly for work, reported that without A11yBoard she would have needed a sighted co-worker or Aira agent to create slides with visual elements like flow charts. She was able to explore, understand, and largely reproduce a complex flow chart slide with shapes, connectors, and text boxes. P2, a graduate student, was able to understand and recreate a slide with five aligned shapes and text boxes—a layout task that was entirely inaccessible through conventional screen readers. Both participants estimated A11yBoard gave them 90-95% independence, noting they would still want sighted confirmation before delivering a presentation but that the tool dramatically reduced back-and-forth with sighted collaborators. Key usability insights included the need for more immediate tactile and audio feedback during editing operations, customizable speech verbosity modes (seven modes were implemented, from brief to absolute pixel values), and the importance of context-aware reporting that emphasizes different properties for different object types. Both participants expressed interest in supplementing touch screen interaction with physical tactile representations like braille displays.
Relevance
This work addresses a critical gap in digital content creation accessibility: while screen readers can navigate the interface elements of presentation software (menus, ribbons, toolbars), they cannot make the 2-D canvas itself accessible. This means blind users have been largely excluded from creating visual content—a significant professional and educational barrier. The five design recommendations offered (provide spatial feedback, tailor feedback to context, offer multimodal creation/editing, consider AI-generated content, balance cognitive load) are broadly applicable to any tool involving accessible 2-D content manipulation, including graphic design tools, whiteboard applications, and data visualization editors. The participatory design approach with a blind co-author throughout the entire process—not just evaluation—demonstrates a model for inclusive research practice. The discussion of AI-generated content (AIGC) as a future direction is particularly forward-looking, noting that while LLMs can generate slides, blind users need accessible tools to review, customize, and control that generated content.
Tags: screen reader accessibility · presentation software · blind and low vision · multimodal interaction · participatory design · content creation · assistive technology · Google Slides