W3C home > Mailing lists > Public > www-voice@w3.org > January to March 2003

Organization of a prompt script in vxml - ssml

From: Greg FitzPatrick <greg@gregfitzpatrick.com>
Date: Mon, 24 Mar 2003 13:07:45 +0100
To: <www-voice@w3.org>
Message-ID: <PIEFJKBIGENLOLNKJIKFCEFECHAA.greg@gregfitzpatrick.com>

My studio records voice prompts for various customers.  Recently a customer came with a script containing over 1200 voice prompts in word.doc.  Our job was to translate the prompts and record them with a voice talent.
Many of the prompts were made up of frequently reoccurring identical phrases (some phrases occurred in over 50 prompts).  Many of the terms were ambiguous or became so in our translations.    

We suggested to the customer that it would be in their interest to "structure" the prompt script. 

We suggested getting their functions into UML and from there getting to XML.  But UML was out of the question. 

In that case in order to dispel ambiguity in translating, versioning, recording, updating and enhancing portability we suggested an XML structure roughly like this:

	(prompt id="xxx")*
		(phrase id="xxx)*
			( text | term id="xxx" )* 

"term" would hold unique words occurring in the text that had specific or stringent definitions which were collected in another document. 

We thought it would be a good idea to promote the use of VoiceXML.  We had used it on earlier occasions to facilitate customer auditioning of our ongoing recordings.  

The stumbling block for us was the lack of ID attributes.  "form" has a name attribute and "block" has an id attribute, but "paragraph" and "sentence" have none. Nor could we find an appropriate element for our "term" - one which would allow an ID.

I assume that someone has considered that prompts are stored in structured formats such as databases and that there must be a convenient way of migrating from there into VoiceXML instances.  

We envisioned that the prompt script could be originated and maintained in VXML even if the customer was not using VXML as an execution platform, but we have not discovered any work in this direction.

1. Can someone help us out?  

2. And does anyone know of the existence of style sheets that would present VXML prompt scripts in HTML primarily for the sake of reviewing literary (non-audio) content?  

Thank you



Received on Monday, 24 March 2003 07:08:01 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:07:36 UTC