![]() |
KnowBrainer Speech Recognition | ![]() |


|
Topic Title: Auto-Transcribe long audio files Topic Summary: Auto-Transcribe in 9.5 doesn't complete long audio files Created On: 08/01/2008 07:07 PM Status: Post and Reply |
|
![]() |
- star125 | - 08/01/2008 07:07 PM |
![]() |
- Lunis Orcutt | - 08/02/2008 06:33 AM |
![]() |
- star125 | - 08/02/2008 10:26 AM |
![]() |
- Chucker | - 08/02/2008 10:41 AM |
![]() |
- star125 | - 08/03/2008 11:36 PM |
|
|
|
|
DNS 9.5: When I use Auto-Transcribe with some lengthy files recorded from a digital voice recorder, the transcriptions simply stop after a sentence or two and the rest of the recording remains untranscribed. The accuracy of the transcribed portion is ok, but I don't understand why it stops.
------------------------- DNS Preferred 9 |
|
|
|
|
|
|
|
|
The professional versions of NaturallySpeaking include an Auto Transcribe folder but this is not available in DNS Preferred so we should probably rule this out 1st. DNS Preferred will allow you to transcribe a single file at a time but if NaturallySpeaking hears nothing (in other words a pause in the dictation) for several seconds, it will automatically turn off the transcribing process.
-------------------------
|
|
|
|
|
|
|
|
|
Thanks for the reply. This would explain what I've found. What I ended up doing was propping the microphone in front of the computer speakers and playing the file with DNS recording dictation, and it did surprisingly well.
------------------------- DNS Preferred 9 |
|
|
|
|
|
|
|
|
Essentially, Lunis' response is what occurs most of the time with any transcription. DNS awaits approximately 10 to 15 seconds when listening for audio to transcribe. It doesn't matter whether you are using auto transcribe or just transcribing in general from an audio file. Once this timeout point has been reached, DNS concludes that the transcription is finished, and that's all she wrote. This is the problem in 99% of the cases like yours. The other 1% include circumstances where in the audio file contains dictation that DNS considers invalid or not being capable of being transcribed. One thing that you can do in such situations is to re-listened to the audio that you are intending to transcribe. When you find gaps that exceed (for editing purposes and ensuring that transcription begins and ends correctly) longer than 7 seconds, edit them to remove the silence. There are any number of good audio editors out there that will allow you to do this, along with converting files from any format to any format. Back when I was originally testing the SDK for version 6, as well as writing the technical specifications for the Panasonic digital voice recorder of that time, I used to use these audio editors to test this problem out. Chuck Runquist "Life's Rule #1: Once you pull the pin, Mr. Grenade is no longer your friend." (Variant of Murphy's Law - Edward A. Murphy, Jr) -------------------------
|
|
|
|
|
|
|
|
|
I had given thought to editing the recording, but decided that it would be more fuss than what I ended up doing. However, in the future that's what I'll attempt if I have a lengthy file to transcribe. I would probably convert it to a wav file and then edit it with CD Wave, a program I find especially easy to use for editing long recordings. Thanks for the suggestion, and knowing that you've edited files for transcription is an incentive for trying it myself.
------------------------- DNS Preferred 9 |
|
|
|
|
FuseTalk Standard Edition v4.0 - © 1999-2013 FuseTalk™ Inc. All rights reserved.