jakarta-taglibs-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "T. V. Raman" <tvra...@us.ibm.com>
Subject RE: RDC: grammar tags in the state component
Date Thu, 24 Feb 2005 13:29:01 GMT


Mich -- Good points all.

Another thing we may want to consider is providing the initial
grammar as a set of utterance/value-generator tupples ---
athought triggered by your "my home state" example.

"My home state" is a good example of personalziation, where the
"default value" is going to be different on a per-user basis.

So one possibility might be for the user of a rdc:state component
to pass in some kind of EL accessor expression 
that gets evaluated at runtime to retrieve the value that is
appropriate for the current user.




>>>>> "Mitch" == Mitch Warner <mitch.warner@fluencyvoice.com> writes:
    Mitch> Rahul, Without worrying too much about the i18n aspect
    Mitch> right now, initially you should maybe just focus on
    Mitch> allowing people to define their own phrases for this
    Mitch> grammar as Daniel pointed out. So rather than having a
    Mitch> grammar which is hard-coded to accept 'default' or
    Mitch> 'initial', it could be configured to accept 'my home
    Mitch> state' or other such semantically equivalent phrases.
    Mitch> 
    Mitch> It would seem that the best apprach maybe to allow
    Mitch> people to override the in-built initial grammar with
    Mitch> another grammar file defined somewhere in the system
    Mitch> 
    Mitch> You mentioned in a pervious mail the following two
    Mitch> issues:
    Mitch> 
    >>> a) How will configuring the initial grammar per RDC
    >>> instance affect
    Mitch> the usability of the application?
    >>> Before, we had a consistent Voice User Interface (VUI)
    >>> for refering
    Mitch> to the initial values.
    Mitch> 
    Mitch> Leaving aside the i18n issue, in some applications and
    Mitch> components it may not make sense for the user to say
    Mitch> 'default' or 'initial'. The example I gave above was
    Mitch> that you may want to allow the user to say 'my home
    Mitch> state' or 'the state where I live'. Additionally, the
    Mitch> 'default' and 'initial' words may be phonetically
    Mitch> similar to other words in the main grammar, which will
    Mitch> reduce the accuracy of the recognition.
    Mitch> 
    >>> b) If we point to a "bad" initial grammar, the rendered
    >>> VoiceXML page
    Mitch> will fail.
    Mitch> 
    Mitch> That is true, but then you would expect the developer
    Mitch> to be able to identify this fault and fix it. Perhaps
    Mitch> you may want to consider checking that the grammar is
    Mitch> well-formed and valid prior to the creation of the
    Mitch> Vxml
    Mitch> 
    Mitch> 
    Mitch> Mitch
    Mitch> 
    Mitch> -----Original Message----- From: Rahul P Akolkar
    Mitch> [mailto:akolkar@us.ibm.com] Sent: 23 February 2005
    Mitch> 16:42 To: Tag Libraries Developers List Subject: RE:
    Mitch> RDC: grammar tags in the state component
    Mitch> 
    >> I'm not a VUI expert, but the idea of having this initial
    >> grammar seems like a fine idea.  I'd think the biggest
    >> issue is providing enough flexibility so that people can
    >> customize it. It's inevitable that users are going to want
    >> to have control over this.  If
    Mitch> 
    >> for nothing else, localization; but also for changing the
    >> utterances that can be used and also doing regional
    >> tuning.
    Mitch> 
    Mitch> Very true. Looking at the RDC tag library from an i18n
    Mitch> perspective, let me identify the resources that are
    Mitch> locale specific, and the action items: 1) Prompts -
    Mitch> Since we already have the mechanism (and necessary
    Mitch> separation) for specifying the prompts associated with
    Mitch> any RDC instance in terms of a config file, localizing
    Mitch> the prompts is simply providing an appropriate config.
    Mitch> 2) Grammars - The component grammars defined in the
    Mitch> tag files which are static URI references need to be
    Mitch> updated with locale specific URIs read out of resource
    Mitch> bundles.  3) Loose ends - Embedded grammars and error
    Mitch> messages. I think the only instance of an embedded
    Mitch> grammar is the initial grammar, though I'll double
    Mitch> check. Error messages produced in the
    Mitch> org.apache.taglibs.rdc.dm package need to change
    Mitch> similar to (2).
    Mitch> 
    Mitch> I will begin to work on the i18n, starting with (2),
    Mitch> then (3). I'll set things up so the grammar URIs are
    Mitch> locale specific, but I lack the ability to translate
    Mitch> ;-) We will need volunteers to do the actual
    Mitch> translation.
    Mitch> 
    Mitch> 
    >> As a follow up question, do you mind pointing me to the
    >> location within the code that generates this dynamic
    >> grammar?
    Mitch> 
    Mitch> Here:
    Mitch> <rdc-src-dir>/src/org/apache/taglibs/rdc/core/BaseModel.java#populateIni
    Mitch> tialGrammar()
    Mitch> 
    Mitch> As per discussion above, it won't stay there for long.
    Mitch> 
    Mitch> Can you also let me know what fails with the Nuance
    Mitch> engine? Does it require an xmlns attribute on the
    Mitch> embedded grammar? I think I noticed that with the
    Mitch> Cisco browser. At some point, we will have a build
    Mitch> switch, so these "platform adaptations" can be done by
    Mitch> specifying the platform as a build property. Also, do
    Mitch> let me know if you are aware of a good compilation of
    Mitch> platform "quirks" for various platforms. That will be
    Mitch> a good starting point for this exercise.
    Mitch> 
    Mitch> -Rahul
    Mitch> 
    Mitch> 
    Mitch> ________________________________________________________________________
    Mitch> This email has been scanned for all viruses by
    Mitch> Netscalibur Mail Scanner, powered by MessageLabs. For
    Mitch> more information on a proactive email security service
    Mitch> working around the clock, around the globe, visit
    Mitch> http://www.netscalibur.co.uk/scanner/index.html
    Mitch> ________________________________________________________________________
    Mitch> 
    Mitch> ---------------------------------------------------------------------
    Mitch> To unsubscribe, e-mail:
    Mitch> taglibs-dev-unsubscribe@jakarta.apache.org For
    Mitch> additional commands, e-mail:
    Mitch> taglibs-dev-help@jakarta.apache.org

-- 
Best Regards,
--raman
------------------------------------------------------------
T. V. Raman:  PhD (Cornell University)
IBM Research: Human Language Technologies
Architect:    RDC --- Conversational And Multimodal WWW Standards
Phone:        1 (408) 927 2608   T-Line 457-2608
Fax:        1 (408) 927 3012     Cell: 1 650 799 5724
Email:        tvraman@us.ibm.com
WWW:      http://almaden.ibm.com/u/tvraman
AIM:      emacspeak
GPG:          http://www.almaden.ibm.com/cs/people/tvraman/raman-almaden.asc
Snail:        IBM Almaden Research Center,
              650 Harry Road
              San Jose 95120

---------------------------------------------------------------------
To unsubscribe, e-mail: taglibs-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: taglibs-dev-help@jakarta.apache.org


Mime
View raw message