VERY large & dynamic vocabulary but very simple grammar - easy?
Jest Ipher
My employer is asking us to find a voice/speech recognition system
that can handle extremely large and constantly changing vocabulary and
is speaker independant.
The extremely large part of the problem: not too hard right? are any
of these packages (was just reading about Nuance) database size
limited in anyway?
The constantly changing vocabulary part of the problem: this is what
I'm worried about. If the words are kind of non-standard (ie not
dictionary words) like product or chemical or biological names - how
much effort is required for training the system to be speaker
independant?
The good part is there is no need to learn a new grammar at all.
Any suggestions on the best way to go for the system? I'm leaning
towards Nuance but I haven't really researched this too hard.
You may already be able to guess our app - it's related to speech
access to large catalogues that constantly get updated...
Thanks a bunch.
Jest
Felix Gofman
Jest Ipher wrote:
> Hi Folks,
>
> My employer is asking us to find a voice/speech recognition system
> that can handle extremely large and constantly changing vocabulary and
> is speaker independant.
>
> The extremely large part of the problem: not too hard right? are any
> of these packages (was just reading about Nuance) database size
> limited in anyway?
>
The particular feature you are talking about is called "dynamic
grammars"
The only size-related limitation is the loading time - the larger the
db the longer it takes. With reasonably sized databases the loading time
is
insignificant.
>
> The constantly changing vocabulary part of the problem: this is what
> I'm worried about. If the words are kind of non-standard (ie not
> dictionary words) like product or chemical or biological names - how
> much effort is required for training the system to be speaker
> independant?
>
Our system is speaker independent and no training is required for
the new dynamically generated entries. Our dictionary has 200,000
entries. If your word is not found there you can use a tool that's
part of our toolkit to automatically generate prons on the fly.
> The good part is there is no need to learn a new grammar at all.
>
> Any suggestions on the best way to go for the system? I'm leaning
> towards Nuance but I haven't really researched this too hard.
Sorry, can't help you with any suggestions other than Nuance ;-)
To get our toolkit you can visit our website and signup for
Nuance Developer Network. If you are serious about this app
consider attending Nuance Speech University and the technical
track of V-World - speech conference. There is more info here:
http://www.nuanceconference.com/
Good Luck!
>
>
> You may already be able to guess our app - it's related to speech
> access to large catalogues that constantly get updated...
>
> Thanks a bunch.
>
> Jest
Regards,
--------------------------------------------
Felix Gofman
Product Manager, Core Software
Nuance Communications
(650) 847-0000
fgofman@nuance_period_com
****************************************
N U A N C E
Speech. Understanding. Action.
http://www.nuance.com
****************************************
Hubert Crepy - IBM France Speech Research
application. This is speaker-independent, can activate large
vocabularies (thousands of names) that change daily (enterprise
phonebook, not unlike a store catalog).
Check out:
http://www-4.ibm.com/software/speech/enterprise/universalaccess_4
html
for a description of an application (Directory Dialer) based on
this technology, and
http://www-4.ibm.com/software/speech/enterprise/technologies_4.ht
ml
for an entry point about our telephony toolkits.
* Sent from RemarQ http://www.remarq.com The Internet's Discussion Network *
The fastest and easiest way to search and participate in Usenet - Free!
Thomas Eisele
called SpeechPearl.
Jest Ipher wrote:
>
> Hi Folks,
>
> My employer is asking us to find a voice/speech recognition system
> that can handle extremely large and constantly changing vocabulary and
> is speaker independant.
>
> The extremely large part of the problem: not too hard right? are any
> of these packages (was just reading about Nuance) database size
> limited in anyway?
>
> The constantly changing vocabulary part of the problem: this is what
> I'm worried about. If the words are kind of non-standard (ie not
> dictionary words) like product or chemical or biological names - how
> much effort is required for training the system to be speaker
> independant?
>
> The good part is there is no need to learn a new grammar at all.
>
> Any suggestions on the best way to go for the system? I'm leaning
> towards Nuance but I haven't really researched this too hard.
>
> You may already be able to guess our app - it's related to speech
> access to large catalogues that constantly get updated...
>
> Thanks a bunch.
>
> Jest
--
+----------------------------------------------------------------------+
| Thomas Eisele Email: Thomas...@philips.com |
| Department Head, Speech Recognition |
| Philips Research Labs Aachen (Germany) Tel.: +49 241 6003-504/507 |
| Weisshausstrasse 2, D-52066 Aachen Fax.: +49 241 6003-518 |
+----------------------------------------------------------------------+