In this extra post, Chris Swift, Learning Technologist at the Holyrood Campus, reflects on the progress of a subtitling pilot project launched in May 2019 at the Moray House School of Education and Sport…
The Moray House subtitling pilot project has been aiming to find an efficient model for subtitling curriculum material used by teaching staff here at the School of Education and Sport. The pilot has now reached its halfway point and, in this blog post, we outline a short summary of what we have achieved so far, some of our findings, and what we hope to accomplish in the remaining months of the project.
At the outset of the pilot there was a recognition that whilst video is increasingly used in teaching and learning, the University did not at that point have a widely available system for subtitling curriculum materials. Our aim was to focus solely on teaching material. We recruited three individuals to the team who had experience subtitling media at other large organisations and Universities. This meant we had a good skill set in place at the start, and could also learn from their knowledge and experience; for example, in using different style guides or subtitling software.
We launched the service in May 2019, and raised awareness through school meetings, staff newsletters, and posters on campus. At the outset, we didn’t have an exact idea of what kind of response we would get, and this has been an important learning point from this project.
Firstly, there were courses we already knew about where students were enrolled with a subtitling adjustment on their schedule. In this case, it was relatively easy to identify and provide subtitles.
Secondly, we worked with Student Disability Services to help identify courses where there were students with a schedule who we may not have been aware of. It was then a case of contacting course directors to offer subtitles.
Thirdly, there were courses where an individual tutor requested subtitles for other reasons – large class numbers for example, or just to provide an option for students reviewing course material online.
This touches on another learning point, namely the variety of material we have been asked to subtitle. It has ranged from DVDs to individual video files, MediaHopper videos to found resources on YouTube or other non-University websites. This has raised questions of copyright and responsibility. Do we subtitle the materials or the original media owners? Do we need consent to do this? It also presented a technical challenge. How do we access videos where we might not be the owner? Where do we publish the subtitled version? We’ve been able to find solutions, but it’s perhaps an extra logistical challenge to factor in when thinking about subtitling teaching material.
Another type of material is lecture recordings. At the start of the project we said we would not include lecture recordings as it’s too labour intensive to subtitle every lecture. However, for a teacher, lectures may be the first thing they’d like subtitled. In this case, it is about managing expectations and timescales.
As a rule of thumb, it takes approximately one hour to subtitle 10 minutes of footage, then additional time for proof reading and editing. For example, an hour long lecture would take approximately six hours to subtitle, plus an additional two-three hours to synchronise the subtitles, proof read, and check for errors.
Other factors can be the speed at which a person talks or the language used in a video. We may need to cross check terminology, technical language, or author names in order to get accuracy in the subtitles. This can all add time to the process and it’s something to bear in mind.
In terms of numbers, we have subtitled 63 videos from nine different courses; the shortest at 1min 26 secs, the longest at 2hrs 7 mins.
As for software, we have used a variety of methods to produce the subtitles. One subtitler uses speech recognition software called Dragon Naturally Speaking and WinCaps subtitling software. Another uses a combination of Amara and YouTube’s automatic captioning tool. We are also looking at OtterAI as another option. Both methods have their advantages and disadvantages. WinCaps is a costed service, but YouTube is free. However, there is potentially an ethical concern using YouTube for University material, even if it is only used temporarily to get the automated captions before being deleted.
As yet, we are not at the stage where we can get perfect subtitles from AI. Computer generated subtitles can speed up the process but creating quality subtitles still requires human intervention. In the pilot at Moray House, we have focused solely on teaching material, and in the final stages of the project we hope to use our data and findings to inform the debate and future policy of the University.
This blog post was originally featured on the Holyrood Campus Learning Technology blog.