With levels of captioning increasing around the world, broadcasters and caption suppliers are always looking for ways to increase efficiency and reduce costs. In The future of pre-recorded subtitling(link is external), Hewson notes that the most time-consuming part of captioning is the creation of accurate text. In recent years, speech recognition technology has increasingly been used to caption live programs, however, there are still severe limitations on the process. A captioner (known as a ‘respeaker’) must repeat all the dialogue in a clear, steady voice – while also speaking the punctuation – for the captions to be anything like acceptable quality. Before they can do this, they must also practice with the software so that it becomes accustomed to their voice.

Speech recognition technology will certainly continue to improve, though, and Hewson looks forward to the advent of ‘speaker-independent voice recognition’ which will be sophisticated enough to generate captions for pre-recorded programs. Systems based on this technology will be able to handle multiple voices. They will also be able to distinguish between voices, split captions into discrete segments for each speaker, and automatically assign colours to identify speakers.

The introduction of these systems is still some years away, and while they will lead to greater efficiencies and streamline production, they will never create perfect captions. “They will continue to struggle with situations of unclear or overlapping dialogue, with television programmes that use new slang, and with (infrequently occurring) specialist subjects that require a lot of research,” writes Hewson. Captioners will therefore continue to play an essential part in the process, but will spend less time on generating text while “their roles in quality control and as the arbiter of (human) taste and judgement are to become more vital than ever”.

To see this article go to Media Access Australia.