Semantic Information Processing of Spoken Language How May I Help You sm

1
Semantic Information Processingof Spoken
Language- How May I Help You? sm

• Alicia Abella
• ATT Labs Research
• Florham Park, New Jersey

2
Motivation

• Goal is to provide automated customer services
  via natural spoken dialog.
  
• Natural means what people actually say, rather
  than what wed like them to.
  
• Shift burden from user to machine

3
Why Spoken Language Understanding?

• User interface as the bottleneck to exploiting
  speech and language processing technological
  advancement
  
• Spoken language as focus of this work
• Machine Initiative
• Menus (please say collect, calling card )
• Classes (please say credit card number,
  destination city, person name, date, )
  
• User Initiative
• How may I help you?

4
Stroustrup on programs which communicate with
people

• "... it must cope with that person's whims,
  conventions and seemingly random errors. Trying
  to force the person to behave in a manner more
  suitable for the machine is often (rightly)
  considered offensive.
• from "The C Programming Language(1987) pp. 76

5
A History of Applications

• Department Store Call-Routing (1989-1991)
• Almanac Data Retrieval (1992)
• Airline Travel (1993)
• Multimodal Blocks World (1993)
• Operator Services (1995-9)
• Customer Care (2000)
• Enterprise Customers (2002)

6
How May I Help You? SM

• Prompt is ATT. How may I help you?
• User responds with unconstrained fluent speech
• System recognizes and determines the meaning of
  users speech, then routes the call
  
• Dialog technology enables task completion

HMIHY
. . .
Local
Account Balance
Calling Plans
Unrecognized Number
7
Extracting Meaning from Speech

• Extracting meaning is primary in speech
  understanding systems.
  
• How to quantify the information content of a
  natural language message?
  
• Such theory is crucial to engineering devices
  which understand and act upon such messages.

8
Communication Paradigm

• Goal of communication is to induce the machine
  to
  
• perform some action
• undergo some internal transformation
• Communication is successful if the machine
  responds appropriately
  
• Contrast with traditional communication theory

9
Shannon (1948)

• The fundamental problem of communication is
  that of reproducing at one point either exactly
  or approximately a message selected at another
  point. Frequently the messages have meaning,
  These semantic aspects of communication are
  irrelevant to the engineering problem.

10

Architecture for Natural Spoken Dialog
Voice reply to customer
Speech
Speech

Text-to-SpeechSynthesis
Automatic SpeechRecognition
TTS
ASR
Data
Words to be synthesized
Words spoken
Words
Words
SLU
SLG
Spoken Language Generation
Spoken LanguageUnderstanding
Meaning
DM
Action
Meaning
Dialogue Management
11
Architecture for Natural Spoken Dialog
Play prompt
ASR
Dialog Manager
Spoken Language Understanding
User speech
Language Models
Acoustic Models
Salient Grammar Fragments
Inheritance Hierarchy
12
Technology Component Traits

• Robustness
• ASR
• Large vocabulary 10,000 words
• Dialects (Nationwide deployment)
• Real-time
• SLU
• Tolerance of varied phraseology
• Many ways of saying the same thing
• Similar way of saying different things
• Real-time
• Dialog
• Confirmation, Re-prompting, Context switching
• Say anything, anytime, anyway

13
Examples of Customer Utterances
Account Balance Other General Billing Rates and
Calling Plans Charge on Bill Change Customer In
fo
14
Comparative Example Unrecognized Number
Customer Care IVR
HMIHY
Sparkle Tone Thank you for calling ATT
Sparkle Tone ATT, How may I help you?
Network Menu
Account Verification Routine
LEC Misdirect Announcement
Account Verification Routine
Main Menu
Reverse Directory Routine
LD Sub-Menu
15
Example Dialogs

• Rate Plan
• Account Balance
• Local Service
• Unrecognized Number
• Threshold Billing
• Billing Credit

16
The technology is in use today

• ATT Customer Care organizations
• Consumer service live since Nov. 2000
• Decreased servicing costs reduced time spent in
  automation reduced repeat calls and customer
  defections
  
• Small Business service pilot underway
• Supports 800s used for billing inquiries and
  corporate calling card transactions.
  
• Determines caller intent to perform activities
  such as making payments, requesting bill
  adjustments, ordering cards, reporting stolen
  cards
• ATT Enterprise customers
• Several in beta trials
• Delivers this functionality in a networked,
  managed environment.
  

17
Industry Specific Applications
Verify coverage, Inquire about a claim, Check
claim status
Check account balances, Apply for mortgage,
Request credit report, Locate branch
Get store hours, Locate nearest store,
Directions, Check inventory availability and
order status
Benefit enrollment, Get a referral, Obtain test
results, Pre-admissions procedures
Obtain a price quote, Make reservations, Get a
seat assignment, Check flight status, Redeem
miles
Get instructions, Report a problem, Obtain
problem status, Order
And Applications Common to All Industries
Password Reset, PIN Reset, FAQs,
Help Desk, Locator Services, Order Entry and
Status
18
Key Value Determinants

• Enhanced Customer
• Experience
• Reduced wait and call times
• even at peak times
• Natural efficient and
• personalized dialog
• Properly fulfills/routes
• request the first time

• Decrease Costs
• Increased use of automation
• reduces servicing costs
• Liberate agent headcount
• Reduce handling time and
• hang-ups and call-backs

• Enhanced Business
• Results
• Implement new applications
• to drive revenue
• No capital investment to
• build or support
• Dynamic, customizable
• resource sharing in secure
• environment

19
Observed Benefits

• Increased automation
• Improved routing
• Ability to add more functionality
• Improved customer satisfaction
• Decreased repeat calls (37)
• Decreased customer defection rate (18)
• Decreased rep time per call (10)
• Decreased customer complaints (78)

20
Research References at www.research.att.com/algor
/hmihye.g. IEEE Computer Magazine, April 2002