Speech recognition is something of a computing holy grail, and has been for a while now. I’m not aware of any commercially available software that will take an MP3 recording of an interview and produce an accurate transcript.On the desktop,
Dragon Naturally Speaking seems best of breed, but this still needs to be trained to recognise the operator’s voice, and, despite the name, is only really competent at dictation rather than transcribing a natural conversation. That said, their premium products are supposed to work with certified digital audio recorders. I don’t think the University provide copies of this.
The equivalent ipad/iphone app is much better, as the audio is beamed to Dragon’s servers for conversion – it would be possible to play this a recording of an interview and see what it comes up with, although there’d be some necessary editing. Amazingly, the app is still free.
Our friends at Google and Apple are working on this, both with the mobile market in mind – Google’s voice search and talk and Apple’s Siri are all rather impressive and don’t require training, but I can’t see any way to get them to do automated transcription. Interestingly, Google will provide automatically generated captions for YouTube videos, (click on the cc button at the bottom on the video of your choice), but my experience suggests these are very far from accurate and I can see no way to export these.
It’s worth questioning whether large scale transcription is actually necessary – at least some of the current qualitative analysis software allows coding of native audio and video, which might offer some advantages in preserving more of the original data.Share