AxsJAX: A Talking Translation Bot Using Google IM: Bringing Web-2.0 Applications to Life
Charles L. Chen, T. V. Raman · 2008 · Proceedings of the 2008 International Cross-Disciplinary Conference on Web Accessibility (W4A) · doi:10.1145/1368044.1368056
Summary
This paper from Google describes how the AxsJAX framework uses WAI-ARIA live regions to make Google Talk — an instant messaging client integrated into GMail — fully accessible to screen reader and self-voicing browser users, and demonstrates a compelling mashup application: a talking multilingual translator. The authors frame the accessibility challenge of Web 2.0 applications through two core concepts. "Reflection" is the ability for adaptive technologies to discover the role and state of interaction widgets, achieved through ARIA role and state properties that make HTML/JavaScript widgets behave like native platform widgets to assistive technology. "Notification" is the ability to detect and communicate relevant page updates to users, achieved through ARIA live regions that let developers identify which portions of a dynamically updating page should be spoken and at what priority. The paper demonstrates these concepts through Google Talk's chat interface within GMail, where incoming messages need to be spoken immediately upon arrival. The technical implementation tags the span element containing the message body (but not the sender name) with live=rude to trigger immediate speech, creating an efficient auditory interface that avoids redundant information in two-way conversations.
Key findings
The paper's most innovative contribution is demonstrating how accessible Web 2.0 components can be composed into new accessible applications — the "mashup" pattern. By combining Google Talk (with ARIA live regions), Google's machine translation bots (accessible via chat addresses like en2zh@bot.talk.google.com), and text-to-speech with proper lang attributes, the authors created a talking translation interface where a user types English and hears Mandarin Chinese spoken back. This required no additional accessibility work beyond the ARIA annotations already in place — the translation bot's responses simply arrived as chat messages with appropriate language attributes, and the existing live region infrastructure automatically spoke them in the correct language. The paper characterizes ARIA primitives as the "assembly language" for Web 2.0 accessibility, with AxsJAX providing higher-level design patterns better suited for web developers. The framework served a dual purpose: delivering immediate end-user value through early-access accessibility solutions, while simultaneously validating and improving the ARIA specification itself through real-world usage.
Relevance
This paper is significant both for its practical contributions and its conceptual framework. The reflection/notification distinction provides a clear mental model for understanding web application accessibility that remains useful today: making individual widgets accessible (reflection via ARIA roles/states) is necessary but insufficient — applications must also communicate dynamic state changes (notification via live regions) in a prioritized, non-overwhelming way. The selective use of live=rude on message content but not sender names demonstrates sophisticated auditory UI design thinking that goes beyond basic compliance. The mashup demonstration — where accessible components compose into accessible applications — illustrates a powerful principle: if individual Web 2.0 services are built with proper ARIA support, their combinations can be accessible "for free." T. V. Raman, one of the authors and a blind computer scientist, brought unique expertise as both a developer and a user of these technologies, having previously created Emacspeak. The AxsJAX framework, while specific to its era, established patterns for ARIA usage that influenced how modern web applications handle dynamic content accessibility.
Tags: ARIA · web accessibility · screen readers · live regions · Web 2.0 · AJAX · dynamic content · multilingual accessibility · assistive technology · instant messaging · open source
Standards referenced: WAI-ARIA