Towards Automating Accessibility in Digital Authoring Workflows

Abstract

Most digital content today remains inaccessible to people with disabilities, who make up 16% of the global population. A long-standing challenge in accessible computing is ensuring digital authors consistently provide the metadata required to make their content accessible through assistive technologies. Despite numerous specialized accessibility tools, authors often lack the time, training, or incentive to use them effectively.

Recent advancements in AI bring an unprecedented opportunity for automating the production and preservation of accessibility metadata in the authoring process. This thesis argues that developers and content creators can use AI tools to make their digital content accessible with minimal disruption to their existing workflows.

It presents two novel AI authoring tools aimed at improving the digital accessibility of websites and scientific documents. The first, CodeA11y, is an AI coding assistant that helps frontend developers produce accessibility-compliant user interface code. The second, WYSIWYM tagging, preserves semantic metadata already present in authored source documents to generate tags and preserve reading order in PDFs.

Through comprehensive evaluations, this thesis demonstrates the effectiveness of these tools and represents a first step toward automating accessibility in mainstream digital authoring workflows.