![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Just tried using the sound recorder function in Audio Commander to dictate my notes for " the Elves and the Shoemaker " and then transcribe it with Dragon. This should be a fairly easy test, as all I'm doing is reading aloud existing handwritten notes. On the other hand, I had to stop on occasions to actually try work what my handwritten notes said. :-)
3 1/2 pages in a reporter's notebook, which will be around 500 words with my handwriting. It took about four minutes to dictate, and 2 or 3 minutes to transcribe, and about 15 minutes to do the corrections... that's rather annoying, particularly as the high error rate is actually higher than I'm achieving most of the time now with dictating blog posts. On the other hand, it will get better as Dragon hears more of my fiction, and part of the problem is that my current computer is pretty much as low a spec as you can get away with and still run Dragon at all. So it's a trade off between speed of recognition and accuracy of recognition, and the install wizard quite rightly went for speed over accuracy. Speed is more important for live dictation, but as it's irrelevant for transcribing a recording, I might fiddle with the settings when I'm doing transcribing.
Dragon does have its own sound recorder built in, which you can use while doing corrections to play back what you said, but in the lower editions this will hold the sound file only until you close the DragonPad document/Dragon itself (still haven't quite worked out which it is). You need to cough up for Professional or its variants to get the facility to save the .wav file permanently so you can transcribe it later.
Conclusions from a brief experiment, then:
The sound file from Audio Commander is usable with Dragon, and at the default sampling rate of 11025 Hz. On my elderly laptop the recognition error rate is pretty high, but certainly acceptable for copy typing in situations where I might not be able to type. It would also be usable for jotting down notes and scraps of dialogue, at least if I could get over my horror at having to listen to my own voice on playback. This might sound daft if you've never tried it, but a couple of my writer friends have also mentioned having the same sort of problem with it. It is actually very disconcerting to listen to your own voice on playback if you're not used to it, as it inevitably sounds different the way you normally hear yourself.
I've saved both the raw transcript and the corrected version as separate files, and might put them online in Google Docs at some point if anyone wants to do a compare and contrast -- it's a useful demonstration of both the types of error and the number of errors.
3 1/2 pages in a reporter's notebook, which will be around 500 words with my handwriting. It took about four minutes to dictate, and 2 or 3 minutes to transcribe, and about 15 minutes to do the corrections... that's rather annoying, particularly as the high error rate is actually higher than I'm achieving most of the time now with dictating blog posts. On the other hand, it will get better as Dragon hears more of my fiction, and part of the problem is that my current computer is pretty much as low a spec as you can get away with and still run Dragon at all. So it's a trade off between speed of recognition and accuracy of recognition, and the install wizard quite rightly went for speed over accuracy. Speed is more important for live dictation, but as it's irrelevant for transcribing a recording, I might fiddle with the settings when I'm doing transcribing.
Dragon does have its own sound recorder built in, which you can use while doing corrections to play back what you said, but in the lower editions this will hold the sound file only until you close the DragonPad document/Dragon itself (still haven't quite worked out which it is). You need to cough up for Professional or its variants to get the facility to save the .wav file permanently so you can transcribe it later.
Conclusions from a brief experiment, then:
The sound file from Audio Commander is usable with Dragon, and at the default sampling rate of 11025 Hz. On my elderly laptop the recognition error rate is pretty high, but certainly acceptable for copy typing in situations where I might not be able to type. It would also be usable for jotting down notes and scraps of dialogue, at least if I could get over my horror at having to listen to my own voice on playback. This might sound daft if you've never tried it, but a couple of my writer friends have also mentioned having the same sort of problem with it. It is actually very disconcerting to listen to your own voice on playback if you're not used to it, as it inevitably sounds different the way you normally hear yourself.
I've saved both the raw transcript and the corrected version as separate files, and might put them online in Google Docs at some point if anyone wants to do a compare and contrast -- it's a useful demonstration of both the types of error and the number of errors.
