Re: [vtp-dev] Problem with PlayPrompt before OptionSet

["Brendan Sterne" <SterneB@xxxxxxxxx>]

Randy et al,

I was looking at the VoiceXML spec again and it seems to me that this
pattern of a form/block/prompt, followed by a separate menu or field will
work if *DTMF* barges in on the prompt, but not if *voice* barges in,
because according to section 4.1.8 of the spec: 

"While in the transitioning state no speech input is collected, accepted
or interpreted. However, DTMF input (including timing information) should
be collected and buffered in the transition state."

As I see it, the voice utterance that barges in on the prompt should not
be accepted during the transition to the menu or field.  Does this
actually work for voice bargein on the Prophecy platform with Nuance?

I can see how this pattern works fine for DTMF because DTMF digits are
collected and buffered in the transition.

Any thoughts / comments are appreciated.

Cheers,
- Brendan

--
Brendan Sterne
Interact Inc Austin R&D
sterneb@xxxxxxxxx


-----Original Message-----
From: Randy Childers <randy@xxxxxxxxxxxxxxx>
To: Voice Tools general developers <vtp-dev@xxxxxxxxxxx>
Date: Mon, 27 Oct 2008 09:05:15 -0500
Subject: Re: [vtp-dev] Problem with PlayPrompt before OptionSet

> Hi Brendan--
> 
> I've not seen this problem before, but then I generally work with
> Nuance and
> Prophecy.  One thing I wasn't clear on:  in the initial Play block,
> where
> you've got caller's barging, is Barge-in active for that block?  I
> would
> expect the symptoms you're describing if Barge-in was set to false, but
> with
> Barge-in set to true, it tends to work.  Also, do you have any prompts
> configured in the Option Set as well?  And what is the Barge-in setting
> for
> the Option Set?
> 
> Randy Childers
> OpenMethods 
> 
> 
> On 10/23/08 1:17 PM, "Brendan Sterne" <SterneB@xxxxxxxxx> wrote:
> 
> > Greetings,
> > 
> > The OptionSet node allows you to configure a grammar for each option.
>  But
> > what if you want a single high-level grammar for the whole OptionSet
> (i.e.
> > "Would you like pizza or spaghetti?").  The pre-built Menu Dialog
> seems to
> > handle this by having a PlayPrompt before the OptionSet. However, I
> am
> > experiencing a problem with this architecture.  If the user speaks
> during
> > the PlayPrompt (barges in), no detection is active, because the voice
> > browser isn't yet processing the OptionSet (<menu>).  When the
> OptionSet
> > (<menu>) starts up, the user is already mid-utterance.  In our case
> we are
> > using Loquendo ASR which complains about trying to recognize
> mid-utterance:
> > Recognizer Error: "API lasrxRecog in module Loquendo ASR 7.5.2.1
> returned
> > with code -134: User has spoken too soon. Utterance started before
> > recognition   The Voice Detector Module has detected that speech has
> begun
> > too early.     Sudden energy fall; maybe the user spoke too early. "
> > 
> > Has anyone else seen this problem?  What is the suggested approach?
> > 
> > Cheers,
> > - Brendan