A Semantic-Web Based Framework for Developing Applications to Improve Accessibility in the WWW
Christos Kouroupetroglou, Michail Salampasis, Athanasios Manitsaris · 2006 · Proceedings of the 2006 International Cross-Disciplinary Workshop on Web Accessibility (W4A) · doi:10.1145/1133219.1133238
Summary
This paper presents a Semantic Web application framework designed to improve web accessibility by adding a metadata annotation layer on top of existing web pages. The framework addresses a core problem for blind users: screen readers serialize web pages linearly, stripping away the visual layout cues that sighted users rely on for rapid page scanning and navigation. Rather than requiring web authors to fix their pages, the framework creates a parallel community that annotates existing pages with semantic descriptions of their structural and navigational elements. The architecture has three key components: ONAR, a graphical ontology editor that allows vocabulary creators to define OWL-based concepts describing web page elements (menus, main content areas, search boxes, navigation links, headers, advertisements, etc.); an annotation tool that lets annotators select DOM elements on web pages and assign ontological concepts to them using RDF metadata stored on a central server; and SeeBrowser, a voice web browser that uses these annotations to provide blind users with browsing shortcuts and page overviews. The framework promotes a community model with four interdependent roles: ontology creators who build vocabularies, annotators who apply concepts to pages, user-agent developers who build tools, and end-users (people with disabilities) who provide feedback. Annotations use XPath expressions to identify DOM elements and are stored on annotation servers inspired by the W3C Annotea project, using Jena API with MySQL for RDF storage and RDQL for querying.
Key findings
The SeeBrowser voice browser was evaluated with blind users who found the annotation-based shortcuts particularly valuable. When visiting an annotated page, users press Alt+I to hear an overview of annotated elements (main content, search box, login form, menu, etc.), then navigate directly to desired elements using Alt+Up/Down arrows. This simulates the visual scanning process that sighted users perform instinctively. Log analysis from user testing showed that annotation-related keystrokes (Alt+I, Alt+Up, Alt+Down) were used in almost every page visited, indicating strong adoption. Users found annotations more helpful on complex real-world sites (like encyclopedias) than on the simple experimental site. The "contains" relationship in the ontology enabled hierarchical navigation — for example, a user hears that a menu exists, and only upon reaching it can drill into individual menu items, preventing information overload. Users requested additional features including "history browsing" (returning to the exact paragraph/sentence position when navigating back), a notepad for collecting information fragments during research, and better table handling. A key limitation was the dependency on human annotators; the paper discusses semi-automatic annotation using templates (one annotation covering multiple pages with similar layouts via regex URL matching) and potential for fully automatic annotation through web structure mining, though acknowledging this remains imperfect.
Relevance
This framework anticipates several important developments in web accessibility. The concept of adding a semantic annotation layer over existing inaccessible web pages prefigures browser extensions and overlay technologies that attempt to improve accessibility without modifying source code — though the community-driven, ontology-based approach here is more principled than commercial overlays. The four-role community model (ontology creators, annotators, developers, end-users) provides a thoughtful governance structure for crowdsourced accessibility improvement. The browsing shortcut feature — providing a page overview that simulates visual scanning — addresses a fundamental challenge that remains relevant: blind users lack the ability to quickly scan and assess page structure that sighted users take for granted. While WAI-ARIA landmarks have since addressed some of these needs natively, many pages still lack proper landmarks, and the annotation approach could supplement them. The work also demonstrates the tension between relying on web authors to create accessible content versus building external tools to compensate for inaccessible authoring — a tension that persists in debates about accessibility overlays and remediation tools.
Tags: semantic web · ontology · web annotation · blind users · voice browser · screen readers · RDF · OWL · information seeking · crowdsourcing · web accessibility
Standards referenced: RDF · OWL