I recently managed a project with the University of Edinburgh to improve their provision of subtitles on their video and audio materials. They have many thousands of hours of recorded content and to comply with legislation, all 'public bodies' should now provide subtitles as standard on their websites. This poses a challenge; how to meet their obligations and provide the best experience for deaf and hard of hearing students or others who access their media, with a limited budget? Accuracy of automated subtitles is hugely variable.
An academic institution with a huge body of international students and staff, as well as a vast array of taught subjects can be a particular challenge for any 'robot' turning speech into subtitles.
We tested a new model to try to address the problem; we employed and trained a group of ten students as subtitling editors to work alongside machine technology to produce accurate subtitles. We subtitled over 53 hours of content in the three month pilot.
The pilot was a success, and a permanent service is now being planned as a result.
I loved working with this motivated and capable group - a mixture of academic, media and disability staff and professional, highly skilled and personable students.
You can find out more about what we did, including a downloadable report using the link below.