OSRA 2002

1
OSRA 2002
The Voice Activated Services Revolution
John Tadlock Principal Member of Technical
Staff Speech and Language Technology SBC
Technology Resources Inc.
2
Memorable Quote
... My greatest discovery of all was
thediscovery of what people wanted to
use. Thomas Edison
3
1878
ATT launches telephone service in New Haven
Connecticut
An operator was required on every call.
And it took 60 years ...
4
1938
... For Rotary Dial telephonesto became dominant.
40 years later ...
5
1978
... Touch Tone had become the telephone interface
of choice.
6
1998
Some 20 years later,the GUI Mouse have became
the norm for accessing information services.
7
A Prediction ... by 2008
60 years
40
20
10
1878
1938
1978
1998
2008
People will use speech to access many new
communications information services.
8
What does it take?
SpokenLanguage Processing
User
New hardware New software New technology An
existing infrastructure
Ultimate System
9
Example Flight Information Service
Spoken Language Processing
Mobile or PSTN Network 18002235436
10
Demo Flight Information
System Tell me just the flight number. Or say I
dont know. User I dont know. System Whats
the departure city? User Austin, Texas. System
Next tell me the arrival city? User St. Louis,
Missouri . System Thanks. Now is that for
departure or arrival info? User Arrival
info. System Tell me only the approximate
arrival time? User 1230 PM System Let me check
that
11
Spoken Language ProcessingA Peek Under the Hood
SpeechRecognition
Language Understanding
Dialog Management
SpeechSynthesis
LanguageGeneration
Dialog Modeling
Knowledge Representation
Language Modeling
Monitoring and Data Collection
12
WhySpeech and Language Technology?

• Enables businesses to control customer care costs
  while increasing customer convenience.
• Opens the World Wide Web (WWW) to the mass
  market.
• Moves communication and information services
  closer to the user as opposed to making the
  customer adapt to the technology.

13
Customer Care Applications
. . . In a survey of 100 call center managers,
speech recognition was most often cited as their
primary need. Frost Sullivan - August 2001
Why?

• Call Volumes are Increasing 24 x 7 access
  required
• Live Agents are Expensive!
• Customer Satisfaction Reduces Hold Times for
  Live Agents
• Competitive Differentiation Customer
  Perceptions
• Consistent Customer Treatment

14
Directory Assistance
Spoken Language Processing
Speech Centric Directory Assistance Database
Existing DirectoryAssistance Database
411
Complex Listings routed to an Operator
SBC took around 1.2 billion directory assistance
callsduring 2001.
Operator Position
15
Demo Directory Assistance
System Thank you for calling Southwestern
Bell. System What city? User St.
Louis. System Say the last name. User Humperdink
. System Say the first name. User John. System
Is that John Humperdink? User Yes. System The
phone number for John Humperdink is314-555-1234.
16
Voice PortalsAlternative Access to the WWW
830 million landline phones 560 million mobile
phonesgrowing to 1.2 billion by 2007
465 millionInternet users

• With over 2 billion landline and mobile users by
  2007spoken language processing enables WWW
  accessvia VXML-based VOICE BROWERS.

Sources Nielsen IDC
17
What is a Voice Portal?
Voice Portal
User
Platform
Voice Web Browser
VXMLInterpreter
Content Providers (Serving VXML documents)
18
Demo Browsing the Voice Web
Voyager Welcome to SBC Voyager. Where
to? User Whats available. Voyager Heres
whats available Your Voice Dial directory
User Portal Services. Voyager Here are the
Portal Services I can take you to User Stock
Quotes. Conversation with the Voice Quotes
site. User Voyager. Voyager What can I do for
you?
User List my bookmarks. Voyager You have four
books marks. Here is the first one User Avis
Confirmation. Conversation with the Avis
site. Conversation with the AMEX Travel
Services site. User Voyager. Voyager What can I
do for you? User Weather
19
Business Opportunities! Jobs!
Platform Vendors Infrastructure Vendors Computer
Vendors
Internet ServiceProviders
Voice Portal Service Providers
Voice Portal
User
Platform
Voice Web Browser
VXMLInterpreter
TelecommunicationService Providers
VoiceBrowserDevelopers
Content Providers (Serving VXML documents)
Web Site Developers And Hosting
Speech And Language UnderstandingVendors
Speech Synthesis Vendors
20
Voice Web vs. WWWHow Fast Will it Grow?

• Faster
• Potentially more customers
• Cost savings for customer care
• New service opportunities
• Need to catch upby voice enabling

• Slower
• Remember Thomas Edison
• How do you make money on the web?
• Designing voice web sites is harder
• Lack of a skilled workforce

21
The Future ... Voice Tone
of calls placed
via a voice portal
via manual dialing
Over the years
At some point customers may elect for off hook
to default to their voice portal service provider
instead of dial tone.
22
OSRA 2002
THANK YOU!
John Tadlock Principal Member of Technical
Staff Speech and Language Technology SBC
Technology Resources Inc.
23
Service Development Process
Effort
Activities
Feasibility /RequirementsAnalysis
Human Factors Performance Analysis
10
Design
20
Dialog Design
Implementation Testing
40
Grammar Build/Test
Data Collection Dialog/Grammar Tuning
30
Data Collection Tuning
24
Another Perspective

• 1. Wizard of OZ studies
• 2. Linguistic data collection and analysis
• 3. Preliminary service prototypes
• 4. User studies data collection
• 5. Field trials data collection
• 6. Initial deployment data collection

25
California Highway Information Network
State of California
Mobile Network PSTN
Highway Status Information
California Highway Traffic Database