Around 10 years ago I led a programme of research looking into the quality of television subtitles. The work was very successful and had a huge impact on the industry, overturning the research on subtitling practices which had led to a poor quality experience for the audience.
If anything our work was too successful, leading to complaints from several academics and eventually a reorganisation at Ofcom, and it was closed down after just 4 years – https://www.bbc.co.uk/rd/projects/live-subtitle-quality
However, taking voluntary redundancy has given me the opportunity to develop new ideas and revisit unfinished business. One piece of unfinished business was the idea that I could automatically monitor the various qualities of broadcast subtitles on a 24/7 basis. I have now developed proof-of-concept software which can measure subtitle delay and word omission on a minute-by-minute basis. I am in the process of uncovering historical trends and current issues which badly affect the quality of subtitles. This work demonstrates the viability of subtitle monitoring and quality control.
Early results indicate that problems are commonplace in both archive programmes and live broadcasts. I have found examples of live subtitles arriving over a minute late and well over a minute early, along with parts of programmes where around half the spoken words are missing from the subtitles. I have now had a paper accepted for presentation at the IBC2025 Technical Conference which will outline the way the software works. I am now looking to find funding to develop this work further and am looking for opportunities to publish other aspects of this work.
My presentation at IBC2025 will be in the session AI in – Speech on Saturday 13 September from
The UK Subtitling Audiences Network is currently seeking funding for this and other aspects of their work and I am open to offers of work related to this research.
