W3C home > Mailing lists > Public > www-voice@w3.org > April to June 2002

Re: ASR without underlying grammar?

From: Mukul Jain <mujain@cisco.com>
Date: Fri, 28 Jun 2002 08:48:08 -0700
To: <www-voice@w3.org>
Message-ID: <PEEDKONLPOPOPJHEJDABMEGPDIAA.mujain@cisco.com>

Hi,

I think this issue is more related to speech recognition servers than to
VoiceXML.

It can be achieved by adding additional platform specific built-in types in
system and reference them using VoiceXML. So It may not be required to add a
additional built-in type. It would be even harder to come up with a correct
built-in grammar given that these phonetic alphabet may not be popularly
used as an standard by everyone.

A platform specific built-in type grammar which could support such a thing
could easily treat the result as if it was uttered properly. I think most
speech recognition engine have support for implementing it.

Thanks,
Mukul Jain
mujain@cisco.com

Date: Thu, 27 Jun 2002 13:01:16 +0200
Subject: ASR without underlying grammar?


Hi,

as far as I understood the specification, there is no possibility to have
ASR without specifying an active grammar in VoiceXML. What if the author
cannot provide a grammar, in cases where the input can't be constrained?
(e.g. when giving proper names) Does VoiceXML deal with that problem?
As a solution, the user could spell the name out, using a built-in grammar
for a phonetic alphabet ("Andrew, Benjamin, Charlie" etc., or anything
equivalent for a specific language), and the Voice Browser could
nevertheless treat the result as if it was uttered properly.
Regards
Tobias Goebel,-
Received on Friday, 28 June 2002 11:48:40 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 30 October 2006 12:48:55 GMT