I'm considering a project that involves tape recordings of railroaders long since gone. I've listened to all the tapes and manually summarized topics and locations on the copy CD. This greatly improved the accessibility to information covered during the interviews, but it's not as fast as reading text. The next step, transcribing the interviews in their entirety, is not particularly appealing in spite of the fact I can type relatively fast.

It occurred to me that voice recognition software may be just the thing to get a rough draft in CorelWP or MSWord. However, after reading several reviews, the software is licensed for one user and apparently will recognize the user's voice and no other. As a result, playing many different tapes and having the software transcribe the spoken words to text may not be feasible, in spite of the fact that the technology would likely do a passable job. Does anyone here have any experience with this techology as an aid to oral/aural history documentation?

Heh... I once worked for a fellow who wanted me to do that very thing. I wound up trying 2 or 3 packages.

All the software is designed to be used interactively. You must train it with your voice. First by you reading it a dozen paragraphs of text the computer gave you to read... and second by using it interactively with your own material, stopping it at mistakes. Of course, while you were training it, it was training you - to enunciate in a way it could better understand. If you didn't stop it on errors, it would snowball, messing up on the next half sentence or so until it recovered.

It had a feature to transcribe recordings you'd already made. But this was obviously a kluge. Obviously it had to know the voice. Then the recorder had to be VERY high quality - ordinary $49 digital voice recorders didn't cut it, it required a pricey $300 one. And then this feature didn't work very well because it wasn't interactive, so errors would snowball.

We wound up doing it ourselves by hand. It took 3-4 hours per hour of recording, using a transcription machine with a foot pedal. Without a foot pedal, it took longer.

Each of the packages allowed as many people to use it as cared to set up a profile. However it couldn't transcribe a conversation between two of them, at any time it was only tuned for the person whose profile was active at the time. This isn't a case of holding back technology... it's just not there.

It remains that the best transcriber is a human ear. Now how much does THAT cost? Go to this site http://www.mturk.com/ This is a site that lets you farm out "micro-tasks" for micro-pay. Now search for "transcription". I see where one outfit chops long recordings into 5-minute bits and pays 75 cents per job to transcribe them.

Thanks for the reply! During the past few days, I found the same thing. You have to train the software with one voice, and it won't handle anything put a near-perfect recording. Since the material I would have to deal with is largely a conversational format of two or more people, I'll have to scratch the software idea. Maybe someday...

Thanks again for replying!