Personalized Assistive Web for Improving Mobile Web Browsing and Accessibility for Visually Impaired Users
Dongsong Zhang, Lina Zhou, Judith O. Uchidiuno, Isil Y. Kilic · 2017 · ACM Transactions on Accessible Computing · doi:10.1145/3053733
Summary
This paper addresses a critical gap in mobile web accessibility: while screen readers enable blind users to access web content, they do not effectively support skimming—the ability to quickly scan content to find relevant information. The authors propose PAW (Personalized Assistive Web), a system with two complementary adaptations. First, a hierarchical outline view automatically transforms any website into a navigable tree structure by parsing HTML headings and sections, presenting only first-level nodes initially with expandable child nodes. This reduces the cognitive load of understanding site structure and allows users to skip irrelevant sections entirely. Second, a personalization layer tracks which sections users expand and access over time, using an algorithm that weights frequency, duration, and recency of access to identify user interests. The top three predicted interest areas appear in a "Quick Links" section at the outline's top. PAW uses a proxy-based three-tier architecture where adaptation happens server-side, making it independent of specific devices, browsers, or screen readers. Each node in the outline receives a numeric index (1, 1.1, 1.2, etc.) that screen readers announce, helping users understand structural relationships and remember locations for return visits.
Key findings
A controlled experiment with 21 blind and 34 sighted participants produced strong evidence for the hierarchical outline's effectiveness. Search time dropped dramatically: blind participants averaged 169.7 seconds without adaptation versus 89.4 seconds with outline view and 60.7 seconds with PAW. Sighted participants showed similar patterns (43.1s → 24.6s → 11.5s). Critically, search accuracy remained consistent across conditions, meaning faster performance did not come at the cost of correctness. Both groups rated the adapted systems higher for perceived ease of use. However, the effects of personalization differed between groups: sighted participants showed statistically significant improvement from personalization, while blind participants' improvement was not significant. Video analysis revealed why—many blind participants started browsing from the middle or bottom of the outline rather than the top where Quick Links appeared, a behavior pattern distinct from sighted users who consistently started at the top. Blind participants particularly valued the numeric indexing scheme, using indices like "2.3" to remember and return to previously found content.
Relevance
This research offers practical insights for mobile web accessibility. The hierarchical outline approach is implementable via proxy servers without requiring website modifications, making it deployable for any site regardless of developer cooperation. The finding that blind users navigate differently from sighted users—not always top-to-bottom—has important implications for personalization design; simply placing important content at the top may not work for screen reader users. The numeric indexing feature emerged as unexpectedly valuable for orientation and should be considered in any outline-based navigation system. For practitioners, the study validates that web adaptation can significantly improve accessibility even when original sites do not follow WCAG guidelines, providing a remediation path for inaccessible legacy content. The substantial sample size (21 blind participants) strengthens the generalizability of findings compared to typical accessibility studies.
Tags: mobile accessibility · screen readers · web adaptation · personalization · blindness · information retrieval
Standards referenced: WCAG · UAAG 2.0 · WAI-ARIA · Section 508