T2S and ASR synergies

[frank.of...@caseris.de]

Hello Arsen and all of you,

we are still trying to find the perfect interface for our MRCP Client.
Therefore it would be nice, if you could explain what synergies a ASR and
a T2S application can have (and how).

I found something like "killonbargein" on a the T2S-channel.
But how is this T2S-channel being told about the ASR SpeechDetected event?
How to do I have to configure 2 channels to make clear, that they belong
together.

Does the T2S/ASR server itself handles the synchronization?

Do I have to create 2 channels and combine them, or do I have to make one
universal channel/resource who is able to handle both?

Many thanks in advance!

Best regards,

Frank Offermanns

[Arsen Chaloyan]

Hello Frank,

On Mon, Nov 16, 2009 at 7:44 PM, <frank.of...@caseris.de> wrote:
>
> Hello Arsen and all of you,
>
> we are still trying to find the perfect interface for our MRCP Client.
> Therefore it would be nice, if you could explain what synergies a ASR and
> a T2S application can have (and how).

Typically an ASR and a TTS channel should be created in the scope of
the same MRCP session.

>
> I  found something like "killonbargein" on a the T2S-channel.
> But how is this T2S-channel being told about the ASR SpeechDetected event?
> How to do I have to configure 2 channels to make clear, that they belong
> together.

The call flow is quite straightforward and the only actual
interconnection is in termination of barge-in-able prompt based on
start of input event. See the following sequence. It assumes both
synthesizer and recognizer resources/channels have already been
allocated.

C->S: SPEAK [synth]
C->S: RECOGNIZE [recog]
S->C: IN-PROGRESS [synth]
S->C: IN-PROGRESS [recog]

S->C: START-OF-INPUT [recog]
C->S: BARGE-IN-OCCURRED [synth]
S->C: COMPLETE [synth]

S->C: RECOGNITION-COMPLETE [recog]

Basically that's it.

>
> Does the T2S/ASR server itself handles the synchronization?

No, it's the responsibility of your application to synchronize them.
Server can be optimized for a quicker kill-on-barge-in response, even
though your application MUST react the same way on START-OF-INPUT
event by invoking BARGE-IN-OCCURRED method.

>
> Do I have to create 2 channels and combine them, or do I have to make one
> universal channel/resource who is able to handle both?

You should create 2 channels anyway. There is no such
resource/channel, which could do that, at least specification provides
no such method.

>
> Many thanks in advance!
>
> Best regards,
>
> Frank Offermanns
>
>
> >
>

--
Arsen Chaloyan
The author of UniMRCP
http://www.unimrcp.org