RE: Question about VoiceXML Implementation Report and speech recognition

Hi Andras,

Sorry to take so long to respond.

We would be very happy to receive your DTMF implementation report
covering 60% of the VoiceXML 2.0 testsuite. We welcome partial reports
like this, so don't feel that you have to cover 100% of the tests. Of
the tests which use 'speech input and speech grammars', then I would
suggest two alternatives:

(a) If you generate an error.noresource error when encountering speech
grammars, then the test will fail. However, you should indicate in the
test report that you are a DTMF only browser, and that you generated
error.noresource when running the test.

(b) In configuring the testsuite for your browser, you need to write an
xslt script which fits your platform (i.e. I assume you have already
modified the sample irtest.xslt to fit your platform running in
automatic mode). You could modify the definitions of conf:grammar and
conf:speech to use DTMF grammars and DTMF input. Note that the tests
themselves should require NO changes, just the stylesheet. If you make
this type change - using DTMF instead of speech - I suggest that you
indicate clearly in your implementation report that you have made this
mode change and which tests are affected. 

We are looking forward to receiving your implementation report.

Thanks

Scott
   

-----Original Message-----
From: Andras Micsik [mailto:micsik@dsd.sztaki.hu] 
Sent: Thursday, March 27, 2003 17:16
To: www-voice@w3.org
Subject: Question about VoiceXML Implementation Report and speech
recognition




Hello,

   our project develops an open source voice browser, and we are in the
process of creating an implementation report for VoiceXML 2.0. Our voice
browser handles most of VoiceXML 2.0, except speech recognition. We
think that VoiceXML is very usable even without speech recognition, and
multilinguality in Europe and Asia (we have Asian partner as well) puts
an extra burden on adding speech recognition to voice browsers.

Unfortunately, the test suite checks many features of VoiceXML (e.g. the
correct implementation of FIA) through the use of speech input and
speech grammars. This means that cca. 40% of the tests contains speech
grammar, which in many cases could be substituted with a DTMF grammar,
without loosing the test target (examples are assertions 143 and 1071).

Our question is whether we can modify these tests to use DTMF
input/grammar instead of speech input/grammar (where appropriate), and
provide an implementation report using these tests?

Best regards,

	Andras Micsik
	PublicVoiceXML project
	W3C Hungarian Office

Received on Friday, 28 March 2003 09:57:07 UTC