Ameliorating the quality issues in live subtitling

Access services have evolved significantly over the past 30 years, and optional subtitles are widely available on mainstream channels in the UK and Europe. Live subtitles now routinely accompany news, sports and chat-shows. The production and consumption of live subtitles both impose heavy cognitive loads, not helped by the constraints of time, practical limitations and the inevitability of errors. Live subtitles of broadcast quality are normally created by real-time transcription of phonetic key strokes or re-spoken text produced by a human intermediary. If the subject matter and vocabulary of the content is not known in advance, transcription errors are very likely. Such errors are distracting or confusing – regulators deprecate them, the press mocks them, but producers have to risk them to deliver the service. Less obvious quality issues also arise to do with the timing lag and the style and position of subtitle text. Recent studies into audience perceptions of live subtitle quality are reviewed, and the results of a pilot study in classifying apparent errors according to likely cause are used to illustrate possible opportunities for mitigation. This suggests that aspects of re-speaking style may be adjusted to enhance accuracy, and that there may be opportunities for new approaches to underpin further quality improvements in future