Accessibility and Automatic Transcription
Online learning can be particularly appropriate for students with many kinds of disability, but care is needed to make sure that they can take full advantage of what it has to offer.
These ideas of accessibility were discussed at RIDE 2020 in a presentation by Tharindu Liyanagunawardena from the online-only University College of Estate Management, London. She began by quoting an anecdote from her time at Reading University in the early days of FutureLearn, where she taught on a beginners’ course in programming. The students were expected to construct a mobile game, and she was surprised to get an email from a student who couldn’t see a red ball on its green background. That student, of course, was colour-blind, and the fix – switching the colours – was very simple. Guidelines for colour combinations that suit students with common types of colour-blindness have now been incorporated into Web Content Accessibility Guidelines to ensure that web content is accessible to people with many types of disability.
Few disabilities can be catered for by as easy a fix as a colour change, however. Deaf students clearly need transcripts of audio and video materials, and these are slow and often costly to produce manually. Automatic transcription is faster and cheaper, with many platforms offering a basic service free, but there are problems with its accuracy: interestingly, Zoom, which has become extraordinarily popular in the coronavirus lockdown, is thought to produce some of the most accurate transcripts.
Tharindu presented the results of a small study at her college in which 283 students and 27 tutors rated the automatic transcripts of a series of webinars on different topics. A large majority of the students and tutors, and almost all international students without English as a first language, found these transcripts helpful despite a fairly low reported average word-for-word accuracy of only 73.3%. A student with dyslexia was particularly enthusiastic about them. Most students and tutors who had no additional needs reported that they found the transcripts useful if the lecturer had spoken too fast or with an unfamiliar accent, or if the audio quality had been poor. Options to download the transcripts in different formats were valued particularly by those with poor connections, who could choose the format that generated the smallest files.
As Tharindu admitted, this was a small and, in some ways, quite limited study, with few participants with any declared disability. The transcripts were least successful in the more specialist disciplines, suggesting that, as expected, the software had difficulty transcribing key technical terms. There might also be a drop in enthusiasm once the novelty factor has worn off. Nevertheless, the study suggested that students and tutors will be prepared and even happy to accept automatic transcripts that are ‘good enough’ rather than perfect. And, as so often happens, an innovation that was designed to improve accessibility for disabled users has been found to benefit the whole user community.