W3C home > Mailing lists > Public > www-voice@w3.org > January to March 2004

Mixed initiative - complete solution - pattern?

From: Royles, Chris <Chris.Royles@vicorp.com>
Date: Thu, 18 Mar 2004 15:17:07 -0000
Message-ID: <A10DE5600B42CB4EABF446C98C053B8E677A4E@Bulldog.vicorp.co.uk>
To: "'www-voice@w3.org'" <www-voice@w3.org>

Hi All,

Apologies, this is an implementation related question but it is relevant to
how VoiceXML is used and its direction in future iterations.

Many voice services that I build demand the use of mixed initiative dialogs.
These dialogs provide the collection of related information, provide
confirmation, and disambiguation. I have already canvassed many examples but
many are fragments, incomplete or contain large volumes of
ECMA-script/JSP/ASP scripting.

There must be a defined pattern for achieving this form of complex dialog in
as pure VoiceXML as possible. I am currently in the process of designing
such a pattern and I am hoping people on the forum can contribute. The
outcome will be a generic and complete example of collecting multiple
related fields, using mixed initiative and disambiguation on multi-slot

A very basic example is the collection of a persons name (two related items,
although it could be many more)

First Name

We could expect utterances as follows

Joe Bloggs
Joe Smith
Jane Doe
Fred Doe
Mr Smith

We would expect an interaction of the form

Example 1
System - 'Please say a name'
Caller - 'Joe'
System - 'I know of two people called Joe, Joe Bloggs and Joe Smith', who
would you like to speak with?
Caller - 'Joe Bloggs'
System - Transferring your call.

Example 2
System 'Please say a name'
Caller - 'Smithe'
System - I didn`t quite get that, did you say 'Smith'?
Caller - 'Yes'
System - Transferring you to 'Joe Smith'

There are a lot of elements that come together to define such a solution,
The grammar differs between the form and the subsequent disambiguation
stage. The re-prompt requires the list of possible matches, there is the
requirement to determine confirmation and accuracy thresholds. How would
such a dialog be written in a straight-forward and extendable manner, so
that we can add such an example to a library of re-usable VoiceXML

With this in mind, are there steps being taken in future iterations of the
VoiceXML standard, to simplify the processing of such 'deceptively simple'
mixed initiative dialogs.

regards, Chris
Received on Thursday, 18 March 2004 10:20:15 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:07:37 UTC