Do Text Transcoders Improve Usability for Disabled Users?
Giorgio Brajnik, Daniela Cancila, Daniela Nicoli, Mery Pignatelli · 2005 · Proceedings of the 2005 International Cross-Disciplinary Workshop on Web Accessibility (W4A) · doi:10.1145/1061811.1061814
Summary
This paper presents the first scientific user study examining whether text transcoders — proxy-based systems that dynamically transform web pages into text-only versions — actually improve usability for disabled users. Text transcoders strip images, JavaScript, multimedia, and layout tables from pages, replacing them with linearized text content enhanced with CSS for liquid layout and resizable text. The study used the LIFT Text Transcoder (LTT) configured with custom annotations to add accessibility features to an Italian government website that lacked alt text, proper headings, form labels, and skip navigation. The experiment involved 29 participants (17 blind, 7 low-vision, 5 with motor disabilities) performing information-finding tasks under two conditions: with and without the transcoder. Participants used their own assistive technologies (19 used screen readers, 4 used magnifiers, 1 used modified mouse/keyboard). The study measured effectiveness (subjective ease of finding information, completion level, error rates), productivity (pages visited, task completion time), and satisfaction through post-task questionnaires on 5-point Likert scales. A pilot study with 11 participants preceded the main experiment to identify methodological issues.
Key findings
The transcoder significantly improved usability across multiple measures, particularly for simple tasks. For effectiveness: task completion level improved from a median of 0.5 to 1.0 (simple tasks, p < 0.003); participants who gave up dropped from 51% to 24% on complex tasks (p < 0.030); wrongly visited pages decreased from 2.59 to 0.97 for simple tasks (p < 0.023). For productivity: visited pages dropped from 4.59 to 2.93 for simple tasks (p < 0.004); task completion time decreased from 465 to 301 seconds for simple tasks (p < 0.047); back button clicks dropped from 1.93 to 0.59 for simple tasks (p < 0.012). For satisfaction: subjective ease of finding information improved from 3.5 to 2.2 (simple tasks, p < 0.001); over 70% preferred the transcoder for future visits. However, the transcoder showed no significant improvement for complex tasks or the sense of knowing where one was within the site, likely because the annotations did not fundamentally alter the site's information architecture. Expert screen reader users benefited least since they could navigate the original site's barriers, while users of outdated assistive technologies (e.g., JAWS 3.5) benefited most as the transcoder enabled tasks they could not otherwise complete.
Relevance
This study is historically significant as rigorous empirical evidence in the ongoing debate about text-only alternatives and, by extension, modern accessibility overlays. The finding that transcoded pages genuinely improved usability challenges blanket dismissals of intermediary accessibility solutions, while the nuances — improvements primarily for simple tasks, no benefit for expert AT users, inability to fix poor information architecture — provide important caveats. The authors correctly identify that the same improvements could be achieved by retrofitting accessibility directly into the original site, but argue transcoders serve two practical roles: as a temporary solution for inaccessible sites and as an alternative interface that may better suit some users. This framing directly anticipates today's debates about accessibility overlay products. For practitioners, the key lessons are: automated remediation can address surface-level barriers (missing alt text, headings, labels, layout linearization) but cannot fix structural problems (poor information architecture, complex interaction design); and user testing with disabled participants reveals that different disability groups experience the same pages very differently, making broad generalizations risky.
Tags: text transcoding · usability testing · accessibility overlays · screen readers · low vision · motor disability · content adaptation · user study
Standards referenced: WCAG 1.0 · Section 508