WebAnywhere: A Screen Reader On-the-Go
Jeffrey P. Bigham, Craig M. Prince, Richard E. Ladner · 2008 · Proceedings of the 2008 International Cross-Disciplinary Conference on Web Accessibility (W4A 2008) · doi:10.1145/1368044.1368060
Summary
This paper introduces WebAnywhere, a web-based, self-voicing screen reader that enables blind users to access the web from virtually any computer with an Internet connection and sound output, without installing any software. The system addresses a fundamental equity problem: while sighted people can use the web from any computer they encounter — library terminals, Internet cafes, friends's laptops, university labs — blind users are restricted to computers with expensive screen readers installed (JAWS and Window-Eyes cost over $1,000 each). WebAnywhere runs entirely within a standard web browser using a client-server architecture. The client-side JavaScript traverses the DOM of loaded web pages using depth-first search, speaking each element's textual representation. Server-side components handle text-to-speech generation using the Festival TTS engine, converting output to small MP3 files that are cached on both server and client. A web proxy fetches and transforms pages to overcome same-origin policy restrictions. The system implements over 30 keyboard shortcuts for navigation by element, heading, link, input field, sentence, word, and character, and supports table navigation by row and column. Users can customize shortcuts and emulate either JAWS or Window-Eyes interaction patterns. Unlike traditional screen readers that maintain an off-screen model, WebAnywhere accesses the DOM directly, which gives it an advantage in handling dynamic Web 2.0 content updates.
Key findings
A user evaluation with 8 blind participants (ages 18-51, ranging from beginner to expert screen reader users) demonstrated that all participants could complete four representative tasks: finding a restaurant phone number on Google, reading a Gmail message, checking a bus schedule, and completing a web survey. Participants agreed there was a strong need for a low-cost, portable screen reader (median rating 5/5) and frequently found themselves at computers without screen readers (median 5/5). Most found WebAnywhere adequate for web access, though slightly more tedious than their usual screen reader. A survey of 15 public computer terminals across Seattle (libraries, Internet cafes, college labs, business centers, a gym, a retirement community) found that 14 out of 15 could run WebAnywhere, despite none having a screen reader installed. Only 2 locations provided headphones, though speakers were available at 5 others. The key technical challenge was latency from server-side speech generation, which was mitigated through aggressive prefetching of upcoming content and caching of common sounds (individual letters as small as 3KB). A significant design insight was that WebAnywhere did not need separate "forms mode" and "browse mode" like traditional screen readers, since direct DOM access eliminated the need for an off-screen model.
Relevance
WebAnywhere represents an important milestone in thinking about assistive technology access as an equity issue, not just a technical one. The paper's core insight — that the cost and installation requirements of screen readers create barriers beyond the disability itself — anticipated broader conversations about assistive technology affordability that continue today. While WebAnywhere itself has been largely superseded by built-in screen readers (VoiceOver on Mac/iOS, TalkBack on Android, Narrator on Windows, and especially the free NVDA), the problems it identified remain relevant: public terminals and shared computers still rarely have screen readers configured, and the digital divide for blind users in developing countries persists. The user-driven design process, where blind consultants shaped feature priorities, is a model for participatory assistive technology development. For web developers, the paper also highlights WebAnywhere's value as an accessibility evaluation tool — a free, instant way to experience a website as a screen reader user.
Tags: screen readers · blind users · assistive technology · web-based · text-to-speech · public access · low cost · user study