Accessibility Commons: A Metadata Infrastructure for Web Accessibility
Shinya Kawanaka, Yevgen Borodin, Jeffrey P. Bigham, Darren Lunn, Hironobu Takagi, Chieko Asakawa · 2008 · Proceedings of the 10th International ACM SIGACCESS Conference on Computers and Accessibility (Assets '08) · doi:10.1145/1414471.1414500
Summary
This paper introduces Accessibility Commons (AC), a shared metadata infrastructure designed to integrate, store, and share accessibility metadata produced by diverse research projects, assistive technologies, and individual users. The authors identify a core problem: numerous projects generate valuable metadata to improve web accessibility for visually impaired users — alternative text for images, heading annotations, content labels, transcoding instructions — but this metadata is siloed in separate tools, stored in incompatible formats, and inaccessible to other systems. The paper surveys eight existing projects and screen readers (aiBrowser, HearSay, WebInSight, Site-wide Annotation, Social Accessibility, AxsJAX, Accessmonkey, and JAWS) to understand what kinds of metadata they produce and consume. From this analysis, the authors identify four common metadata components: URI addressing (which pages the metadata applies to), element addressing (which HTML elements it targets), conditions (when the metadata is applicable), and semantics (what the metadata means). They distinguish between declarative metadata (static annotations like alternative text and headings) and procedural metadata (executable scripts like those in Accessmonkey). The AC infrastructure consists of a flexible database schema that can accommodate these varied metadata types and a server-side conversion system that transforms metadata between formats so that applications can consume metadata originally produced by incompatible tools. A prototype was implemented using Ruby on Rails with a MySQL backend.
Key findings
The survey of existing tools revealed that while projects use different URI addressing methods (wildcards, regular expressions, domain matching), element addressing techniques (XPath, URI, MD5 hashing), and condition types, they could all be mapped to a common four-component schema. The prototype demonstrated successful metadata integration across three case studies: aiBrowser metadata was separated into reusable general-purpose components and application-specific procedural components; HearSay migrated its metadata storage to the AC schema, enabling label sharing across a wider user base; and SADIe contributed large-scale site-wide metadata using CSS and OWL-based structural analysis. The server-side metadata conversion system could transform between formats — for example, converting wildcard URIs to regular expressions, or image URIs to XPath expressions — enabling cross-application metadata reuse. The prototype was deployed as the backend for the Social Accessibility project. The authors identified key challenges including metadata conflicts (multiple alternative texts for the same image), spam prevention, and database performance at scale, noting that domain-based indexing would help manage the expected millions of metadata records.
Relevance
This paper represents an early and ambitious attempt to address a problem that persists today: accessibility improvements created by one tool or community remain locked within that ecosystem. The concept of a shared accessibility metadata commons anticipated later developments in crowdsourced accessibility (like WebAIM community efforts) and social approaches to web repair. For practitioners, the paper highlights the value of standardized, shareable accessibility annotations — the same principle behind modern standards like WAI-ARIA and schema.org accessibility properties. The distinction between declarative and procedural metadata remains relevant when considering which accessibility fixes can be automated versus which require human judgment. While the specific infrastructure proposed here did not become a universal standard, its vision of collaborative, cross-tool accessibility improvement continues to influence research on social accessibility, crowdsourced auditing, and browser-based accessibility overlays.
Tags: web accessibility · metadata · assistive technology · screen readers · social accessibility · crowdsourcing · alternative text · web transcoding
Standards referenced: WAI-ARIA · WCAG · RDF · OWL · XPath · CSS