VoiceXML for Developers

1

• VoiceXML for Developers

Art Clarke Director, Government Solutions Tellme
Networks, Inc Presentation Material Provided by
Scott McGlashan, CTO Pipebeach
2
What is VoiceXML?

• an XML-based markup language exposing Speech,
  Internet, and Telephony resources, enabling the
  creation of conversational dialogs over the
  phone
• Key Design Principles
• Input speech recognition and DTMF
• Output pre-recorded audio and synthesized speech
  
• Internet XML, IP, HTTP, SSL, JavaScript
• Telephony call transfer, data passing

3
Familiar Web Architecture
4
Back-end Integration and VoiceXML
5
W3C Voice Browser Working Group

• Founded May 1999
• 60 company members
• Mission Standards group to prepare and review
  markup languages to enable internet-based speech
  applications
• Publishes requirements and specifications for
  languages in the W3C Speech Interface Framework
• http//www.w3.org/Voice

6
W3C Speech Interface Framework
VoiceXML v2.0
Speech Grammars
Speech Synthesis
Call Control
Semantic Interpretation
7
Road to Standardization

• VoiceXML v1.0 (May 2000)
• VoiceXML Forum
• Specification submitted to the W3C
• VoiceXML v2.0
• W3C Voice Browser Working Group
• 50 members collaborating
• Addressed 400 change requests

8
VoiceXML Forum

• Industry Group to promote VoiceXML
• 550 member companies
• Submitted VoiceXML 1.0 to W3C in May 2000
• Key areas of Forum responsibility
• Education
• Marketing
• Conformance
• http//www.voicexml.org

9
VoiceXML Adoption

• Activities
• W3C (50 members), VoiceXML Forum (550 members)
• Companies
• Tellme Networks, Nuance, NMS, IBM, BeVocal,
  Speechworks, Cisco, Intel, ATT, Cisco, Motorola,
  IBM, Nortel, Lucent, Comverse, Verizon,
  Pipebeach
• Real-world implementations
• ATT, Toll Free Directory Assistance
  (800-555-1212), voice portals, financial
  brokerages, carriers ATT Wireless, BellSouth,
  Cingular, Vodafone

10
Programming VoiceXML

• Writing a VoiceXML application is programming!
• Programming in VoiceXML is a mix of procedural
  programming and declarative programming
• Control constructs are procedural (if-else etc.)
• Forms are mainly declarative the VoiceXML
  platform iterates through a ltformgt until values
  for all field items have been collected (FIA,
  Form Interpretation Algorithm)

11
Document
doc1.vxml

• A VoiceXML document defines one or more dialogs
• The user is always in one dialog at any time
• Each dialog specifies the next dialog to
  transition to using a URL

Dialog 1
Transition dialog 2
Dialog 2
Transition http//xyz.com/doc2.vxml
12
Dialog

• A Dialog describes an interaction between a user
  and the system
• Two kinds of dialogs form and menu

13
Form

• A form defines an interaction that collects
  values for a set of field items. Each field item
  may specify output, input and evaluation

input

• ltformgt
• ltfield name"travellersgt
• ltgrammar modevoice src./number.grxml/gt
• ltpromptgtHow many are travelling?lt/promptgt
• ltfilledgt
• ltsubmit nexthttp//travel.com/order/gt
• lt/filledgt
• lt/fieldgt
• lt/formgt

output
eval
14
Menu
ltmenu idcommandsgt What service would you
like? ltchoice next/carsgt Car hire
lt/choicegt ltchoice next/hotelsgt Hotel
reservations lt/choicegt ltchoice next/newsgt
Todays news lt/choicegt lt/menugt

• A menu is a simplified form
• A menu presents the user with a choice of options
  and then transitions to another dialog based on
  that choice

15
FIA - Form Interpretation Algorithm
Simplified!

• The FIA has a main loop that repeatedly selects a
  form item and then visits it
• The first (in document order) form item, whose
  field item variable is undefined, is selected
• As a result, the user is prompted for each field
  item in turn

16
FIA Form Example

• ltformgt
• ltfield namedest-city"gt
• ltpromptgtWhere do you want to go to?lt/promptgt
• ltgrammar modevoice src./cities.grxml/gt
• lt/fieldgt
• ltfield name"travellersgt
• ltpromptgtHow many are travelling to your
  destination?lt/promptgt
• ltgrammar modevoice src./number.grxml/gt
• lt/fieldgt
• lt!-- other fields --gt
• lt/formgt

field item variable
Field item 1
Field item 2
17
Executable Content

• Executable content is a block of procedural logic
• Executable content gives developer dynamic
  control over system behavior
• Executable content may appear in
• ltblockgt
• ltfilledgt
• event handlers
• Other behavior is controlled by the VoiceXML
  platform, such as Form Interpretation Algorithm

18
Filled Field Items (Evaluation)

• ltformgt
• ltfield namedest-city"gt
• ltpromptgt Where do you want to go to?
  lt/promptgt
• ltgrammar modevoice src./cities.grxml/gt
• lt/fieldgt
• ltfield name"travellersgt
• ltpromptgt How many are travelling to your
  destination? lt/promptgt
• ltgrammar modevoice src./number.grxml/gt
• lt/fieldgt
• ltfilledgt
• lt!-- acknowledge order --gt
• Thank you. Your order is now being
  processed.
• lt/filledgt
• lt/formgt

19
Goto
ltgoto nextdocument2/gt
document2
document1
form1
form1
formItem1
ltgoto nextitemformItem2/gt
formItem2
form2
ltgoto nextform2/gt
20
Submit

• Typically used to send results from client to
  server
• Syntaxltsubmit nextURI namelistvar1 var2
  .../gt

HTTP GET/POST
web server
cgi, servlet etc.
VoiceXML platform
internet
21
Submit, Example

• ltformgt
• ltfield namedest-city"gt
• ltpromptgt Where do you want to go to?
  lt/promptgt
• ltgrammar modevoice src./cities.grxml/gt
• lt/fieldgt
• ltfield name"travellersgt
• ltpromptgt How many are travelling to ltvalue
  expr"city"/gt? lt/promptgt
• ltgrammar modevoice src./number.grxml/gt
• lt/fieldgt
• ltfilledgt
• Thank you. Your order is now being processed.
• ltsubmit next"http//travel.com/order"
  namelistdest-city travellers"/gt
• lt/filledgt
• lt/formgt

22
Scripts

• Scripts are client side programs
• Client-side vs. Server-side processing
• Server-side process/time intensive, accessing
  large/dynamic information resources, etc
• Client-side validate input, simple computation,
  state navigation, etc
• Only ECMAScript can be used for scripting in
  VoiceXML
• VoiceXML platforms must support ECMAScript
• ECMAScript - http//www.ecma.ch/ecma1/STAND/ECMA-2
  62.HTM

23
Scripts, Example

• ltscriptgt
• function factorial(n)
• return (n lt 1)? 1 n factorial(n-1)
• lt/scriptgt
• ltform id"form"gt
• ltfield name"fact"gt
• ltpromptgt Say a number and hear its
  factorial.lt/promptgt
• ltgrammar modevoice src./number.grxml/gt
• ltfilledgt
• ltvalue expr"fact"/gt factorial is
• ltvalue expr"factorial(fact)"/gt
• lt/filledgt
• lt/fieldgt
• lt/formgt

24
if, else and elseif
ltformgt ... ltfilledgt ltif cond"travellers
gt 10"gt Sorry, we cannot handle groups larger
than 10 persons ltclear namelist"travellers"
/gt ltelseif cond"travellers gt 5 dest-city
'London'"/gt Sorry, we cannot handle
groups larger than 5 persons travelling to
London ltclear namelistcity travellers"/gt
ltelse/gt ltsubmit next"http//travel.com/o
rder"/gt lt/ifgt lt/filledgt lt/formgt
25
Variables

• Variables can be manipulated and referenced
• declare ltfield name"user2"gt
• assign ltassign name"user1" exprpeter"/gt
• clear ltclear namelist"user1 user2"/gt
• Reference How many are travelling to ltvalue
  exprdest-city/gt ?
• VoiceXML variables are ECMAScript variables

26
Variable Scope
Session variables are read-only variables
provided by the interpreter context
Scope defined by element containing executable
content (ltblockgt, ltfilledgt or event handler)
Search for variable name
27
Variable Scope, Example
document
dialog

• ltvar name"counter" expr"5"/gt
• ltformgt
• ltvar name"counter" expr"1"/gt lt!-- hides
  document.counter --gt
• ltvar name"counter2" expr"document.counter"/gt
  lt!-- 5 --gt
• lt!-- Other form contents --gt
• lt/formgt

ltformgt ltvar name"counter" expr"1"/gt lt!--
hides document.counter --gt ltvar name"counter2"
expr"document.counter"/gt lt!-- 5 --gt lt!--
Other form contents --gt lt/formgt
Referencing variables in enclosing scopes can be
done by using the scopes logic scope name as
prefix
28
Events

• Events are used to signal unexpected situations
  
• Events are thrown when
• There is a syntax or semantics error in the
  docment
• The user does not provide audio input, or an
  intelligible response or hangs up
• A ltthrowgt element is encountered
• Events are caught by an catch event handler
• ltcatch eventcom.acme.mailreadergt...lt/catchgt
• ltcatch eventnomatch noinputgt...lt/catchgt
• Shortcut ltnomatchgt is equivalent to ltcatch
  event"nomatch"gt
• Other shortcuts ltnoinputgt, lterrorgt

29
Events, Example

• ltfield namedest-city"gt
• ltpromptgt Where do you want to go to?
  lt/promptgt
• ltgrammar modevoice src./cities.grxml/gt
• ltnomatchgt
• Please say the city you want to fly to.
• lt/nomatchgt
• lt/fieldgt

30
Useful Links

• W3C (World Wide Web Consortium)
• http//www.w3.org
• W3C Voice Browser Working Group
• http//www.w3.org/voice
• VoiceXML v2.0 Specification
• http//www.w3.org/TR/2001/WD-voicexml20-20011023/
• VoiceXML Forum
• http//www.voicexmlforum.org

31

• VoiceXML The Future

32
Vision for the Future

• Ubiquitous access to web content through Voice
• Enterprise web infrastructure serving all this
  content in XML (Studio .NET, modperl, PHP3, JSP)
• More natural user interfaces that learn from the
  caller
• New devices and new telephony networks enabling
  the next generation of applications and services

33
Advancements in XML

• Document Object Model
• Methods
• Properties
• Events
• Separation of data and presentation
• XSLT
• Xforms
• loading XML resources
• Modularization
• Schemas
• Namespaces
• Xlink, RDF

34
Advances in Voice User Interfaces

• New recognition technologies
• Statistical Language Models
• Speaker Verification
• Richer semantic processing
• Better ability to intuit caller intent
• Understand where barge in occurs
• Always listening
• Easier concatenated speech
• Iteration constructs to loop over a list of audio
  files
• Cleaner separation of prompt resources for
  internationalization

35
Advances in Core Infrastructure From the Network
to the Device

• VoIP networks with SIP
• In-network call control at different layers
  (CCXML)
• New multi-modal devices -- 3G, telematics

36
Challenges

• Seamless transition to the future
• Resolving the requirements of new devices, all
  XML languages, seamlessly with the requirements
  for a simple but powerful dialog language

37
Good News!

• VoiceXML 2.0 is nearing Last Call Working Draft!
• Customers are building and launching applications
  for millions of callers!
• Agreement on general direction of activity in the
  Voice Browser Working Group!
• Lots of industry activity toward this next
  generation!
• Strong interest in converging all the activities!