Transforming Flash to XML for Accessibility Evaluations
Shin Saito, Hironobu Takagi, Chieko Asakawa · 2006 · Proceedings of the 8th International ACM SIGACCESS Conference on Computers and Accessibility (Assets '06) · doi:10.1145/1168987.1169015
Summary
This paper from IBM Tokyo Research Laboratory proposes a method for transforming Flash content into XML structures to enable automated accessibility evaluation. At the time of publication, Flash was installed on over 95% of personal computers and was a major source of inaccessible web content. While HTML accessibility checking tools were well established, Flash's binary SWF format made it opaque to external inspection tools. The authors developed a system that combines two complementary approaches to extract accessibility information from Flash content: obtaining output through Microsoft Active Accessibility (MSAA) via the Flash Player, and injecting ActionScript bridge code directly into SWF files to access their internal object structures. A Transcoding Proxy intercepts SWF files, injects the AS-JS bridge code, and passes the modified content to a Browser Module. The MSAA bridge and AS-JS bridge each capture different aspects of the content — MSAA provides widget roles, ALT text, and reading order, while the AS bridge reveals the intended content structure, developer-specified accessibility properties, and object positions. The extracted information is converted into an XML DOM that can be evaluated using familiar DOM-based checking methods similar to those used for HTML.
Key findings
The prototype checker was evaluated against 58 existing webpages containing Flash with GUI components. The AS bridge injection succeeded for almost all pages (53 out of 58), while MSAA processing succeeded for 49 pages. Of the 28 webpages where full AS-bridge-based accessibility checking was possible, only one webpage was designed to comply with accessibility guidelines — the Fujitsu Plasma Display page, which was created following Fujitsu's own Web Accessibility Guidelines. Across all evaluated Flash content, only 1.1% of graphics and 23.0% of buttons had alternative text, demonstrating widespread neglect of Flash accessibility. The research also revealed that most developers were unaware of Flash's _accImpl (accessibility implementation) mechanism for assigning proper widget roles, and that the Flash Player's auto-label function often generated misleading text for assistive technology users. The two extraction approaches proved complementary: MSAA could detect static text and screen reader output that the AS bridge could not, while the AS bridge preserved content structure and developer-intended accessibility properties that MSAA flattened.
Relevance
Although Flash is now obsolete, this paper remains relevant as a case study in the recurring challenge of making rich, non-HTML web content accessible. The core problem — proprietary binary formats that resist external accessibility inspection — reappears with each new generation of web technology, from Silverlight to modern JavaScript-heavy single-page applications and canvas-based rendering. The paper's approach of transforming opaque content into a structured, inspectable XML representation anticipates modern accessibility testing strategies that build DOM-based accessibility trees from complex web applications. The finding that only one out of 28 Flash sites followed accessibility guidelines, despite the existence of authoring tools and best practice documents, underscores a persistent truth: guidelines alone are insufficient without robust tooling and developer awareness. The dual-extraction methodology — combining platform accessibility API output with direct content inspection — remains a sound architectural pattern for accessibility evaluation tools today.
Tags: Flash accessibility · automated testing · accessibility evaluation · rich internet applications · screen readers · XML transformation · MSAA
Standards referenced: WCAG 2.0 · Section 508