Workshop on Speaker biometrics and VoiceXML 3.0 — Summary

On 5-6 March 2009, the W3C Voice Browser Working Group held a Workshop on Speaker biometrics and VoiceXML 3.0 in Menlo Park, California, US, hosted by SRI International.

The minutes of the workshop are available on the W3C Web server:
http://www.w3.org/2008/08/siv/minutes.html

There were 16 attendees from the following organizations:

• SRI International
• Recognition Technologies, Inc.
• J. Markowitz, Consultants
• Intervoice
• EIG
• Deutsche Telekom AG, Laboratories
• Centrelink
• iBiometrics, Inc.
• Daon
• Cisco Systems, Inc.
• iBiometrics, Inc.
• Voxeo
• Nuance
• General Motors
• W3C

This workshop was narrowly focused on identifying and prioritizing directions for SIV standards work as a means of making SIV more useful in current and emerging markets. Topics discussed during the workshop includes:

• SIV use cases (Application requirements for SIV in VoiceXML 3.0)
• SIV users (design philosophy, uncertainty, security, identity)
• Audio formats for SIV (Wav, PCM, alaw, ulaw, OGG, etc.)
• Data format for multimodal applications (EMMA, etc.)
• SIV related standards other than W3C (CBEFF, INCITS 456, BIAS, BioAPI)
• SIV and MRCP V2
• Architecture and functionality (features, configuration, APIs, etc.)

During the workshop we have clarified "Why SIV functionality should be added to VoiceXML" as follows.

• The system would be more responsive, so VoiceXML could shorten customer perceived latency and provide performance benefits to the users.
• It would be easier for developers to generate applications, because programming interface would be consistent with the way they use other VoiceXML resources and low-level operations would be hidden to them.
• Adding SIV to a standard would make it portable and facilitate integration with Web model, because it makes SIV applications consistent with the model and provide efficiencies of scale in hosted environment.
• Standardizations of easy to use API would minimize vendor lock-in and grow the market.
• Support in VoiceXML enables SIV use (without the application server) with intermittent/offline connectivity.
• Standards are a sign of technology maturity.

The major "takeaway" is our confirming SIV fits into the VoiceXML space and generating the "Menlo Park Model", an SIV available VoiceXML architecture, as below.

1. Main hidden security issues that people have concern about are idendified, and ways in which they can be realistically addressed are discussed. Those issues don't disappear but now we know we can address them.
2. VoiceXML 3.0 could be an example of the Interaction Managers within the W3C's MMI Architecture. The synchronization and markup integration of multiple modalities should be addressed. There are likely to be multiple modalities/factors involved in an interaction using VoiceXML. Consequently, developers need a way to not completely separate those modalities.
3. Collaboration with other W3C Working Groups and other standard bodies, e.g., OASIS/BIAS, is expected.

Note that the discussion during the workshop was mainly focused on the application side and provided little guidance for engine providers. So a good standard that would create a nice wrapper around speech engines would be needed. Classification, segmentation and identification should be also considered and the group needs to determine whether or not to include them in VoiceXML V3.0.

The Call for Participation, the Logistics, the Presentation Guideline, the Agenda and the Minutes are also available on the W3C Web server.

Judith Markowitz, Ken Rehor and Kazuyuki Ashimura, Workshop co-Chairs

$Id: summary.html,v 1.12 2009/06/23 14:19:55 ashimura Exp $