Tech: Software advice needed

Subject: Tech: Software advice needed
From: Jim Carroll
Date: 05 Mar 08 - 02:58 PM

Can anybody suggest software for PC to convert recorded speech (lectures, interviews etc) into text?
Thanks
Jim Carroll

Subject: RE: Tech: Software advice needed
From: treewind
Date: 05 Mar 08 - 03:15 PM

IBM used to have quite a good package for this. It was designed for dictation of text and commands into a PC via a microphone; I don't know how well suited it was for processing audio files.

Be warned that it's still a very difficult and error prone technology and you will have to do a lot of checking and correcting. If the original soundtrack isn't very clear it may not work at all.

Anahata

Subject: RE: Tech: Software advice needed
From: Bert
Date: 05 Mar 08 - 04:25 PM

You might try Dragon

Subject: RE: Tech: Software advice needed
From: Stilly River Sage
Date: 06 Mar 08 - 12:50 AM

Ask Don Firth. He has posted about voice recognition software to several threads. He might have some ideas about recorded voice with such software.

SRS

Subject: RE: Tech: Software advice needed
From: GUEST,Jon
Date: 06 Mar 08 - 04:54 AM

I'd guess it would be difficult. As far as I know, this type of software requires some training to adapt to a users voice/accent, users may need to be trained how to speak, the software works best in quiet environments, etc.

I might be wrong but everything would seem to me to go against this application.

Subject: RE: Tech: Software advice needed
From: Andrez
Date: 06 Mar 08 - 05:43 AM

I did a lot of work with IBM Via Voice when it first came out and was better supported than it is now. From what I know Dragon Naturally Speaking is the product of choice now but although I was given a copy I havent been too inspired to roadtest it.

Yes you do need to train the software and based on my Via Voice experience accents are a real issue as the bias is towards American English. In the end I found that the time I spent (and I worked hard at it in work settings) wasnt worth the effort as I could simply transcribe the spoken text on a word processsor quicker than the voice software could do it.

Another source of error was in using the headset mike. You can calibrate the software and hardware fairly well but the next time you run a session, you cant get the headset in exactly the same position it was the time before. That meant more time wasted calibrating.

Oh and on top of that the worst things was that there real problems with phonetic howlers. That meant spending more time doing some really close proofreading for contextual errors that spell checkers couldnt pick up.

I'd love to find a program that did the conversion well but I dont think its there yet.


Hope this helps,

Cheers,

Andrez

Subject: RE: Tech: Software advice needed
From: GUEST,doc.tom
Date: 06 Mar 08 - 06:01 AM

Hi Jim,
I had some experience of voice recognition software in a previous existence(and borrowed it). They all seem to be designed for dictation (the Chief Excutive even got rid of his Secretary)- there were phenomenal problems with working with recorded material. It needs training for EACH voice that it tries to interpret - and any extraneous noise (like background) really buggers it up. I'd agree with those who say it's more trouble than it's worth! - Sorry!

TomB