READ: A (Research) Platform for Evaluating Non-visual Access Methods to Digital Documents
Laurent Sorin, Julie Lemarié, Mustapha Mojahid · 2015 · ASSETS '15: Proceedings of the 17th International ACM SIGACCESS Conference on Computers & Accessibility · doi:10.1145/2700648.2811338
Summary
READ (Documents Architecture REstitution) is a configurable research platform for evaluating alternative methods of non-visual access to digital documents. The system addresses a well-documented problem: despite extensive functionality, screen readers remain frustrating and inefficient for blind users, even when accessing content that follows accessibility guidelines. The inherent linearity of audio and tactile modalities is frequently cited as a core limitation. The platform was developed in C# for Windows, using the SAPI (Speech Application Programming Interface) standard for text-to-speech. Its architecture follows a "blocks and events" paradigm: documents are decomposed into information blocks, each containing one or more events that can play sequentially or in parallel. Relationships between blocks—whether visual (spatial layout) or discourse-based (e.g., illustration relationships)—are specified in separate XML files and can drive navigation functions. Configuration files define how document tags map to output events. Any XML vocabulary (HTML, custom tags) can be used as long as the configuration matches. The system supports multiple access "modes" that users can switch between during navigation. Output capabilities include SAPI-compatible text-to-speech with SSML control, WAV audio files for auditory icons, and spatialized sound that positions audio in 3D space. A plugin mechanism allows researchers to extend the system with new input devices, output modalities (such as haptic feedback), or navigation functions without modifying the core application.
Key findings
The platform was validated in a user study with 60 blindfolded and 12 blind participants whose task was to comprehend a 1,500-word structured text. The system proved stable and bug-free across all participants. Usability scores using the System Usability Scale (SUS) reached approximately 90 out of 100 for the best rendering methods, demonstrating that READ can implement highly usable interactions. The paper positions READ as an alternative to modifying open-source screen readers, which have complex architectures due to their need to handle operating system interactions. When research focuses specifically on document accessibility rather than general computer access, READ provides a simpler, more controlled environment. Every user and system action is automatically logged to text files, supporting quantitative analysis of navigation behaviors and reading patterns. The primary limitation acknowledged is that XML configuration files, relationship specifications, and documents must be manually edited. While time-consuming for large documents, the authors consider this acceptable for research purposes where controlled stimuli are typical.
Relevance
READ addresses infrastructure needs in accessibility research. Evaluating novel document access paradigms—such as concurrent speech, aural glancing, multimodal rendering, or spatialized audio—currently requires either building custom tools or modifying complex screen reader codebases. READ offers a middle path: a flexible platform where researchers can implement and compare approaches with minimal development effort. The platform could support research on several open questions: How effective are non-linear navigation strategies compared to sequential reading? Can spatialized audio convey document structure effectively? How do multimodal approaches (combining audio with haptics) affect comprehension and cognitive load? What are the informational equivalents to visual text formatting (bold, headings, indentation) in non-visual modalities? For practitioners, the paper reinforces that guideline compliance alone does not guarantee usability—screen reader access remains "frustrating and poorly efficient" even for compliant content. This argues for continued research on fundamental access paradigms rather than solely focusing on standards compliance. The planned open-source release under Creative Commons licensing would enable broader research community participation.
Tags: document accessibility · research platform · non-visual access · screen readers · text-to-speech · multimodal · auditory interface · spatialized audio · research tools
Standards referenced: SAPI · SSML