At the Voxeo Customer Summit 2010, Dir. of Speech Technologies Dan Burnett provided an update on the evolving VoiceXML 3 standard. More information at: http://www.voxeo.com/ http://www.voxeo.com/summit2010 http://blogs.voxeo.com/speakingofstandards/

1. Standards Update:
VoiceXML 3
Dan Burnett, Ph.D.
Dir. of Speech Technologies, Voxeo
(Dir. of Standards, Voxeo)

2. Voxeo on Standards
 Develop ahead of standards
 Make it Open Source
 Lead in standards creation
 Lead in standards adoption
© Voxeo Corporation

3. Past Leadership
 W3C
• VoiceXML 2.0/2.1, SRGS 1.0, SISR
1.0, SSML 1.0
• CCXML 1.0, SCXML 1.0, EMMA 1.0
 IETF
• MRCPv1 extensions, MRCPv2,
P-charge-info, SIP security
© Voxeo Corporation

4. Where we are now
 W3C
• VoiceXML 3, SSML 1.1, Pronunciation
Alphabet Registry, Speech in HTML 5
• CCXML 1.0, SCXML 1.0, EMMA next, MMI
architecture
 IETF, 3GPP
• MRCPv2, XMPP (incl. multi-party Jingle and
multiple chat), Media Control, SIP Overload,
SIPREC, CODEC (Speex)
 JCP
• JSR 289, 309 – SIP servlets, media control
• JSR 154, 254 – Java servlets and servlet
pages
• XMPP SIP servlet – submitting to JCP
© Voxeo Corporation

5. VoiceXML
VoiceXML
3
VoiceXML
VoiceXML 2.1
2.0
VoiceXML
1.0
2000 2004 2007 2010
© Voxeo Corporation

6. VoiceXML
VoiceXML
3
VoiceXML
VoiceXML 2.1
2.0
VoiceXML
1.0
2000 2004 2007 2010
© Voxeo Corporation

7. V3 Motivations
 FIA flexibility
 New features
 Extensibility
 Better integration with other W3C languages
© Voxeo Corporation

8. V3 is . . .
 a restructured core
 some new features
 convenience elements to mimic VoiceXML 2.1
© Voxeo Corporation

9. V3 Architecture
 Core functionality defined in modules
 Modules combined with convenience syntax into
profiles
© Voxeo Corporation

10. Core functionality defined in modules
 Module behavior defined precisely as state
machines
© Voxeo Corporation

11. Modules + Conv. Syntax = Profiles
 Modules grouped into profiles
 Legacy (V2.1), Basic, Maximal
 Convenience syntax simplifies authoring
© Voxeo Corporation

12. Convenience Syntax
 New elements and attributes, but no new
functionality
 Behavior defined in terms of core functionality
 For example, <menu> defined in terms of
<form> with grammars and prompts
© Voxeo Corporation

13. Convenience Syntax
 Definite candidates are
• menu/choice/enumerate/option
• error/help/noinput/nomatch shortcuts
• link
 Possible (but different) candidates might be
• if/else/elseif (using SCXML)
• transfer (using CCXML)
© Voxeo Corporation

14. New Stuff
 New media, SIV functions
 Session root documents
 Real-time controls
 Author-specifiable transition controllers
 V2 eventing model now async & compatible
with DOM Level 3
© Voxeo Corporation

15. New Functionality – Video
 Video -- <audio> replaced by <media>, which
allows both audio and video
<media type="audio/x-wav" src="http://www.example.com/resource.wav"/>
<media type="video/3gpp" src="http://www.example.com/resource.3gp"/>
<media> <!-- inline SSML with audio media fallback-->
<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis">
Ich bin ein Berliner.
</speak>
<media type="audio/x-wav" src="ichbineinberliner.wav">
</media>
© Voxeo Corporation

16. New Functionality – Media Control
 Media control -- media clipping, speed, and
volume control now possible without resorting
to SSML
<media type="audio/x-wav" soundLevel="+6.0dB" speed="50%" repeatcount= "2"
src="http://www.example.com/resource.wav"/>
<media type="video/3gpp" clipBegin= "2s" clipEnd="5s" repeatDur="25s"
src="http://www.example.com/resource.3gp"/>
© Voxeo Corporation

17. New Functionality – SIV
 SIV – speaker authentication capabilities
available as core functionality
• Enrollment – creates voice model, associates it with
id in speaker database
• Identification – which voice model in speaker
database is a match for the speech?
• Verification – for the claimed id,
does the speech match the voice
model in the speaker database?
© Voxeo Corporation

18. New Control – Session Root
 Just like application root
<vxml session="blahblah.vxml" ...>
 Well, not exactly
• If not specified, no session root
• Session root change is ignored or causes error
 First, let’s review application roots
© Voxeo Corporation

19. Application Root Review
A: <vxml>
AppRoot A
B: <vxml>
AppRoot B
C: <vxml root="B">
AppRoot B
D: <vxml root="E">
AppRoot E
F: <vxml root="E">
AppRoot E
G: <vxml>
AppRoot G
© Voxeo Corporation

20. Session Root
A: <vxml>
No Session Root
B: <vxml session="C">
Session Root C
D: <vxml>
Session Root C
E: <vxml session="F" >
Session Root C
G: <vxml session="H" requiresession="true">
error.badfetch
© Voxeo Corporation

21. Real-time Controls
 Special grammars that are always active (not just in
the wait state)
• Allows arbitrary speech/dtmf
• Immediate: volume, speed, skip
• At next event processing: cancel, goto
<form>
<rtc grammar="digit3.grxml" action="volume" params="+5"/>
<field name="a"> ... </field>
<field name="b">
<cancelrtc grammar= "digit3.grxml "/>
...
</field>
</form>
 Acts as pre-filter on input stream,
replacing matches with silence
© Voxeo Corporation

22. Transition Controllers
 Inter-element transitions now under author
control
 Controllers at form, document, application, and
perhaps session levels
• e.g. form controller specifies which form item to
execute next
 Controllers can be in SCXML or another flow
control language
 Default controllers will give FIA behavior in
Legacy Profile
© Voxeo Corporation

23. Transition Controllers Example 1
<!-- document-level transition controller controls inter-form transitions -->
<vxml ...>
<controller ...>
<scxml:scxml version="1.0" ...>
<!-- SCXML code determining which form to go to next -->
</scxml>
</controller>
<form id="form_a" >
...
<goto next="form_b"/> <!-- goto is only a suggestion now -->
</form>
<form id="form_b" >
...
</form>
...
</vxml>
© Voxeo Corporation

24. Transition Controllers Example 2
<!-- form-level transition controller controls inter-field transitions -->
<vxml ...>
<form>
<controller src= "myformbehavior.scxml">
<field name="field_a" > ... </field>
<field name="field_b" > ... </field>
<field name="field_c" > ... </field>
<field name="field_d" > ... </field>
</form>
...
</vxml>
© Voxeo Corporation

25. For More V3 Info
 Follow the work
• http://www.w3.org/Voice
 Check out our recent Developer Jam Session
• http://developers.voiceobjects.com/tech-topics/
monthly-jam-sessions/
 Contact me
• dburnett at voxeo dot com
Dan Burnett, Ph.D.
Dir. of Speech Technologies, Voxeo
© Voxeo Corporation