Friday, 7 October 2016

Automatically getting from recording to transcript

Usually there is only one way that you can create a transcript of a sound or video file: and that is to type it out yourself.

You would think that you could play a recording of your data to Dragon Naturally Speaking and that it would record it for you. Unfortunately, Dragon requires a high quality sound input, and speakers are not usually high enough quality for it to recognise what is being played.

Additionally, if there is more than one person speaking on your sound file, transcription software simply doesn't work. We should really call it 'dictation' software.

So instead you have to listen to your recording, and then repeat it into Dragon. You edit as you go - and as Dragon misinterprets your accent. Repeatedly.

If you can get funding, pay someone who is skilled at this to do it for you. Then all you need to do is to edit the resultant files for sense.

When I did my Master's, I lost six months of my life to transcription - or avoiding transcription - as it takes about one minute to transcribe each second of data recorded. You can read more about my trails in that space here.

However, there are some other ways. You can change your mp3 audio file into an mp4, upload it to YouTube, and wait for YouTube to create an auto transcript. More instructions on that here

You can then copy out the transcript into Word, tidy it up using Dragon Naturally Speaking.

However, Dragon costs.

And there are a couple of free options.
I haven't tried either of these free options, but they would be worth a go.



No comments :

Post a Comment

Thanks for your feedback. The elves will post it shortly.