Prev: Differentiator with FIR filter
Next: FreeRTOS configTICK_RATE_HZ weird behavior
From: alexryu on 11 Feb 2010 20:10
Can someone point me to a relatively modern and general summary of the
progress of transcribing phone calls (I would like to transcribe many
calls, each with different speakers, and different types of phone (is that
last important?))? Is this even feasible in a business/industrial context?
I've read some academic papers, but I suppose that is not their concern.
Thanks for any guidelines you can give me.
From: alexryu on 11 Feb 2010 20:29
>Can someone point me to a relatively modern and general summary of the
>progress of transcribing phone calls (I would like to transcribe many
>calls, each with different speakers, and different types of phone (is
that
>last important?))? Is this even feasible in a business/industrial
context?
> I've read some academic papers, but I suppose that is not their concern.

>Thanks for any guidelines you can give me.
>
Another thing I am interested in as a possible alternative is keyword
spotting, which to me seems somewhat more tractable than the full
transcription problem. Does anyone have any first hand experience?
From: Tim Wescott on 11 Feb 2010 22:10
On Thu, 11 Feb 2010 19:10:24 -0600, alexryu wrote:

> Can someone point me to a relatively modern and general summary of the
> progress of transcribing phone calls (I would like to transcribe many
> calls, each with different speakers, and different types of phone (is
> that last important?))? Is this even feasible in a business/industrial
> context?
> I've read some academic papers, but I suppose that is not their
> concern.
> Thanks for any guidelines you can give me.

I believe that current state of the art involves routing the call to
Jamaica or India or some other place where folks speak English well and
don't expect to make much money, then pretending that it's done by
machine.

--
www.wescottdesign.com
From: fatalist on 12 Feb 2010 09:06
On Feb 11, 8:10 pm, "alexryu" <ryu.a...(a)gmail.com> wrote:
> Can someone point me to a relatively modern and general summary of the
> progress of transcribing phone calls (I would like to transcribe many
> calls, each with different speakers, and different types of phone (is that
> last important?))?  Is this even feasible in a business/industrial context?
>  I've read some academic papers, but I suppose that is not their concern.
> Thanks for any guidelines you can give me.

Google "spinvox scandal"

You can also try Google voice (a fully-automated voice transcription
engine)
but don't expect to be too impressed
 | 
Pages: 1
Prev: Differentiator with FIR filter
Next: FreeRTOS configTICK_RATE_HZ weird behavior