Accessibility Information Needs in the Enterprise
Sharon Snider, Willie L. Scott II, Shari Trewin · 2020 · ACM Transactions on Accessible Computing (TACCESS) · doi:10.1145/3368620
Summary
This paper from IBM presents the first empirical analysis of accessibility information needs within a large multinational enterprise, analyzing 439 unique accessibility questions and 362 unique keywords gathered over two years from FAQs, support channels, a chatbot, a glossary, and search logs. The research reveals that enterprise employees's most pressing accessibility questions center on testing, delivering (recording and reporting compliance), development, and compliance — practical, process-oriented concerns rather than abstract understanding of accessibility principles. Ten question topics emerged, ranging from specific technical questions ("What is the ARIA role for a navigation element?") to process questions ("How is compliance reported to customers?") to general understanding ("What does accessibility mean?"). Eight types of information were sought: definitions, relations between concepts, procedural steps, resources, tool guidance, deeper understanding of rationale, connections to expert people/teams, and impact of changes. The authors found that up to 66 percent of these questions could potentially be answered by an accessibility ontology, but existing ontologies like ACCESSIBLE covered only 26 percent of the concepts appearing in the questions. To address this gap, they created the Enterprise Accessibility Conformance Ontology (EACO), an open OWL-based ontology with 122 classes and 1,922 individuals that extends the ACCESSIBLE ontology with enterprise-specific concepts around conformance processes, testing workflows, reporting, and design guidance.
Key findings
EACO, combined with ACCESSIBLE, covered 86 percent of the accessibility concepts identified in the enterprise questions, up from 26 percent with ACCESSIBLE alone. When applied to IBM's accessibility chatbot, EACO improved question-answering performance from 69 percent to 91 percent accuracy by enabling the system to match questions to relevant resource articles using ontological concepts rather than exact keyword matching — making it robust to terminology variations (e.g., matching "JAWS" to "screen reader" to "assistive technology"). Validation against a second enterprise's accessibility FAQ showed 64 percent concept coverage, suggesting reasonable generalizability while identifying areas for expansion. A notable finding was the near-total absence of user perspective questions — the data was dominated by standards compliance and testing processes, with very few questions about how people with disabilities actually use technology. The authors acknowledge this may reflect the enterprise's compliance-focused culture or that user-oriented information is more readily available externally. The most common relation questions asked: "What deliverables need to be tested?" (22 questions), "What tool to use for a specific task?" (21), and "What types of deliverable are required to be accessible?" (19).
Relevance
This research is directly valuable for any organization building an enterprise accessibility program. The detailed taxonomy of question topics and information types provides a blueprint for structuring accessibility knowledge bases, FAQs, and training materials. The finding that employees struggle most with process questions — how to test, what to report, which tools to use — rather than understanding what accessibility means suggests that organizations should invest more in practical, task-oriented guidance rather than general awareness content. EACO itself is openly available on GitHub and could be adopted or extended by other organizations to structure their own accessibility knowledge management. The chatbot improvement from 69 to 91 percent demonstrates a concrete, measurable return on investment for ontology-based approaches to accessibility support. For accessibility teams that are perpetually resource-constrained, automating answers to routine conformance questions could free experts to focus on more complex design and evaluation work.
Tags: enterprise accessibility · ontology · knowledge management · accessibility conformance · chatbot · question answering · WCAG · Section 508 · accessibility testing · organizational accessibility
Standards referenced: WCAG 2.0 · Section 508 · EN 301 549 · WAI-ARIA