Training DNS for Unavailable Dictator Options

[DanaPNW]

I am a medical transcriptionist and have been using Dragon Medical V8 for redictation by me of dictation from doctors. I would like to try bypassing the redictation step and run the actual doctor dictation through Dragon and then edit the result. The doctors for which I would use this method speak very clearly and without accent.

I am hung up, though, on how to train Dragon for each of these users. I am a subcontractor and get my work through an outsourcing company, so I have no direct contact with these doctors and therefore cannot ask them to read specific text to train Dragon.

Does anyone out there know of a way to do this? I tried searching the archives, but couldn't find anything.

I record the remotely recorded dictation onto my computer over phone lines via a digital transcriber machine.

Any suggestions/info would be appreciated.

[Coop]

DanaPNW,

Since no one has replied to your question, I will give it a shot. It seems to me that if you truly have:

1. a clear audio sample of the required user (doctor's dictation) and
2. you have in text form the words the doctor spoke (your transcription of the doctor's recording)

It should be possible to:

1. Create a new enrollment (training) text per:

http://www.synapseadaptive.com/joel/changi...llmentatext.htm

2. make a wave file of the doctor's audio sample (recording) and then
3. use the wave file to train a new digital recorder user

This procedure is only slightly different from what is required to train a new user for a digital recorder... with the exception of creating the new enrollment text.

Once you have created the new digital recorder user for your doctor, you will have to teach and maintain it by performing corrections and using the vocabulary builder to learn his style from previously created documents. Since you are a seasoned DNS user already, none of this is new.

Realize, that I have never done this myself but have often been tempted to try. Since you have a real need for this to succeed, you are the perfect guinea pig.

I suspect that the real pitfall(s) of this technique will be getting a clean and clear audio sample (wave file). You indicate that your audio sample comes from a digital dictation machine via the phone lines. I have no way of knowing if this machine is of sufficient quality to adequately train DNS or if the doc uses a good DNS friendly dictation style.

In any event, it's worth a try. Please let us know how it works out. If successful or not many others will benefit from your experience.

Hope this helps,

-Coop

[snoekie]

You have a problem. To try to get them to train may result in loss of work, and if upu correct via keyboard you are likely to end up with bandly corrupted voice file.

My take.

[DanaPNW]

DanaPNW,

Since no one has replied to your question, I will give it a shot.  It seems to me that if you truly have:

1.  a clear audio sample of the required user (doctor's dictation) and
2.  you have in text form the words the doctor spoke (your transcription of the doctor's recording)

It should be possible to:

1.  Create a new enrollment (training) text per:

http://www.synapseadaptive.com/joel/changi...llmentatext.htm

[snipped]

-Coop
[right][snapback]211653[/snapback][/right]

Coop,

Thanks so much for this information! This is EXACTLY what I was thinking of but could not figure out how to do it.

I suspect you're right that the quality of the audio sample may be an issue, as it is quite variable, but hopefully that will not be a barrier.

Once I give this a try, I'll post the results.

Thanks again.