During a panel discussion, Claude Le Guyader from Deluxe Media noted that access suppliers across the world seized on voice recognition as a captioning solution because of the shortage of stenographers. He emphasised though that the technology, which is becoming increasingly common for other applications such as the iPhone’s Siri, is far from perfect.
Voice recognition was first used for captioning on Australian television in 2005. In this method, a captioner ‘respeaks’ the dialogue of a news bulletin or other live program into a microphone, and voice recognition software converts it to text. Media Access Australia’s Chris Mikul said, “Captions created using voice recognition can be of equivalent quality to those done by stenocaptioners, but a lot depends on how much time captioners have spent practising with the software so it gets accustomed to their voices, and how much preparation time they have before each program.”
Voice recognition software is still a long way from being able to accurately transcribe speech without having an intermediary ‘respeaking’ into it – especially when the sound quality is less than perfect. The most notorious example of this is the automatic captioning now available for YouTube videos. Commenting on the CSI conference, the author of the ‘i heart subtitles’ blog wrote, “At the moment this does illustrate the technology’s limitations as most readers here I am sure are aware – the captions created this way are completely inaccurate most of the time and therefore useless.” Nevertheless, if you own the YouTube video, it is now possible to correct the automatic captions.
A full transcript of the conferenceis available on the CSI Magazine website.
Top of page