Microsoft Speech Server

spectacularscarecrowAI and Robotics

Nov 17, 2013 (3 years and 10 months ago)

63 views

Microsoft Speech Server




Charlene Volker


Solutions Sales Specialist

Lynn Olson


Technical Specialist

2

Government Self
-
Service Applications


Permits and
Inspections


Jury Management


Citation Management


Elections


Tax Information


Utilities


Password Reset



Auto Attendant


Outbound Notification


License Application


License Renewal


Motor Vehicle
Renewal


Mobil Workforce


Speech Enabled
Portals


3

State and Local Governments are
looking for IT Solutions That…


Provide a solution for increased services
demands by constituents


Demonstrate a rapid return on investment


Are easy to install and maintain


Address key operational challenges




Microsoft Speech Technologies provide an easy, economical way to
extend government services to a broader range of constituents,
across a wide array of devices.



4

Benefits of Speech Recognition


Reduces Labor Costs in the Enterprise


Call Center, IT and Operations


Automates repetitive high volume, low value tasks


Improves Automation


Simplifies complex transactions


Adds convenient access to existing systems


Improves the User Experience


Speech provides advantages over other methods
including touch
-
tone, email and web


80% believe speech is as easy or better to use than
internet


5

Web

Speech

Where Speech Fits

Accessibility

User
Friendliness


Customers

Reached


Touch
-
Tone

6


7


8


9


New York Department of Education


1.2 million students, 80,000 teachers and 1,200
schools


Highly Acclaimed Website


Low computer literacy and lack of internet access


Digital Divide


Speech
-
enabled ASP.NET website


Extends NYDOE website to phone


Provides attendance, grades, lunch menu and other
key information to parents


NY DOE Chose Speech Server




We have a large number of parents who are not comfortable with a computer. But
telephones are accessible and easy to use…using Microsoft Speech Technologies, parents will
have access to information 24/7.”


Richard Langford, Deputy CIO NYC DOE

10


11


12

Microsoft Speech Technology Vision

Microsoft Developer tools for speech

Contact

Center

IVR

Messaging

Enterprise

& LOB

Applications

Desktop

&

Embedded

Microsoft Speech Server

Speech Technologies

Making Speech Mainstream

“Speech and natural language understanding are the key technologies that will
have the most impact in the next 15 years.”

--

Bill Gates

13

Permits and Inspections


Schedule Inspections


Cancel/Reschedule Inspections


Post Inspection Results


Obtain Inspection Results


Fax Results


Outbound Notification


14

Municipal Court Systems


Check the status of case or citation


Check the status of an account


Amount Owed


Due Date


Payment History


Schedule and/or Confirm


Court Date


Administrative Review


Hearing


Credit card payment processing


Jury Notification


Outbound Court Appearance reminder call

15

Tax Assessment & Collections


Status Checks


Current year tax payments


1
st

half


2
nd

half


Prior Year Tax Payments


By year


Delinquent or lien status


Credit Card Payments of Taxes


Fax Options


Parcel Detail


Property and Ownership detail


Payment History


Billing History


16

Utility Department


Account Inquiry


Current balance


Adjusted balance


Last payment amount


Pending payment


Current billing date


Delinquent due date


Last payment date


Delinquent balance



Start/Stop Service


Payment Site Locator


Outbound Notification


Meter Reading


Outage Reporting
and Status



17

Back End Integration


Microsoft CRM


Accela Advantage
(Tidemark)


Accela Automation


Accela Permits Plus


Black Bear


Computronix


CRW


CSDC Amanda


Echelon3 Technologies



Eden


Hansen


Harris Computer
Systems


KIVA


Municipal CityView


Pentamation


Springbrook Software


SunGard HTE



18

Microsoft’s Auto Attendant


Automated operator


55,000+ names in Microsoft directory


Results


75% first
-
try task completion rate


50% reduction in agent headcount


ROI (over 4 years) = 1400%


Payback period = 4 months


Annual savings = $270,000



What Is Microsoft Speech
Server?

20


Web
-
based, speech
-
enabled IVR &
multimodal server


DTMF and speech


Supports open and industry standards


Integrated ASR/TTS


U.S. English, U.S. Spanish, Canadian French


High performance, scalable, reliable, secure


Flexible deployment options


Comprehensive and robust development
using VS .NET and Speech Application SDK


Reporting, analysis, and tuning tools


Integrated Windows management


Leverages existing IT, Web, telephony, and
data assets and skills

R2

21

More Self
-
Service, Less Cost

Voice
User
Interface

Application

Logic and
Business
Rules

Back
-
end
Data

Graphical
User
Interface


Same
development tools


Same web and
web services
infrastructure


Same application
logic and back
-
end data


Same
management and
monitoring tools

22

Microsoft Speech Server

Media

Gateway

PSTN

Microsoft
®

Speech

Application SDK

+

Visual Studio
®

.NET

Development Workstation


Telephony Card

Telephony

Interface

Manager (TIM)

Speech Engine

Services (SES)

Telephony

Application

Services (TAS)

ASP.NET Web Server

ASP.NET Controls

Grammars

Prompts

Web Pages

Speech

Server

Topology

Telephony

HTML + SALT +
JScript

PBX

TDM

TDM

SQL


Logging &
Reporting

JH


DB

Back
-
end/DB Connectivity

23

MSS Configuration Options


o
All
-
in
-
one
-

manageability

o
Distributed
-

fail
-
over

o
Universal Port: ASR, TTS, and Languages

o
Licensed per Proc

PBX
(optional)

Microsoft Speech
Server and Web Server

Standard Edition

24 ports or less


Microsoft Speech Server

(TAS service enabled)

Web Server

and DB(s)

PBX

Microsoft Speech Server

(SES service enabled)

Load balancer

Enterprise Edition

AIO or Distributed

SOAP

or

or

+1

24

Advantages of Speech


Improve Constituent Satisfaction



Extend Services Easily



Leverage Existing Technology



25

For More Information…


My contact Information


cvolker@microsoft.com



http://www.microsoft.com/speech/solutions
/government/default.mspx



Case Studies


Data Sheet


Sample Application Wav Files


26

© 2003
-
2004 Microsoft Corporation. All rights reserved.

This presentation is for informational purposes only. Microsoft makes no warranties, express or implied, in this summary.

27

Speech Application SDK Overview


VS.NET integrated tools


Dialog


Grammar


Prompts


ASP.NET speech controls


Grammar libraries


Debugging tools


Debugger


Telephony simulator


Log player


Analysis tools


Call analyzer


Reports


Samples, tutorials, docs

28


Design the VUI interaction


Call Flow


Create grammars for expected answers


Add speech controls to a Web Form


Extract the meaning (answers) of the
recognition results and bind them to
controls


Create the prompts


Debug before deploying


Building a Speech Application

Grammar

Editor

Dialog

Editor

Prompt

Editor

Debugging

Tools

29

Grammars

30


A structured list of rules that identify the
words or phrases that can be used for
speech input and subsequent automatic
recognition


Maintained in .grxml files


adhere to the
SRGS format

I would like a

I’ll have a a

Give me a

small

medium

large

plain

cheese

mushroom

ham

pizza

pizza please

Preamble

Postamble

Semantic Items

What are Grammars?

31

Grammar Elements


Phrase (words)


List (alternatives)


Rule reference


Group (rolls Lists up)


Advanced:


Wildcard


Halt


Skip


Script Tag (assigns
variable)

32

Creating Grammars and Semantic Mark
-
up

33

Adding Semantic Mark
-
up


Add Script Tags to your grammar rules


After the phrase where a semantic value is
required


This creates <tag> elements in the underlying
.grxml file


Use the Semantic Script Editor to create script


Create semantic elements and attributes


Use free format script editing for complex
requirements


34

Speech Controls

35

Basic

Speech Controls

Prompt

Speech Controls

Listen

Dialog

Speech Controls

Application

Speech Controls

QA

SemanticMap

SemanticItem

Grammar

Command

AnswerCall

CreditCardNumber

Phone

DataTableNavigator

Custom Controls

CompareValidator /

CustomValidator

Speech Controls

36

What are Application Speech Controls?


Composite controls, to collect
common information









Examples include date, phone number
and credit card controls



Application Control



Dialog Speech Controls

Grammars

Default Prompts

37

What are QA Controls?


A QA manages a question
-
answer interaction with a user


Asks a question (prompts the user)


Accepts voice input (using a
grammar)

qaPSize

qaPTopping

Which size Pizza
would you like to
have?

What toppings
would you like on
your pizza?

Question
-
Answer
Interactions

Application

38


Semantic items store elements of
contextual information gathered from the
user


A semantic map is a container for
multiple semantic items


SemanticMap

QA Control(s)

Answers

Collection

siPSize

Semantic Item State

Empty

Needs Confirmation

siPTopping

Confirmed

Semantic Items and Semantic Map

39

Dialog Control Editor

40

Prompts

41

MSS Prompts


Supports Pre
-
Recorded and Text
-
to
-
Speech


Stored in a Prompt Database


Easy to Manage


Prompt DB and Editor enable you to:


Add Transcriptions


Create Extractions


Record Prompts


Edit Wave Files


Validate Prompt Coverage



42

The Prompt Editor

43

Speech Debugging

44

The Telephony Application Simulator

45

Edit SML Results

Trace Output

Speech Control State

Simulated Exception
Conditions

Break on

Listen Start

Enable SML Editing

Text or Voice Input

Play Prompts

The Speech Debugging Console

46

Managing your MSS Deployment


Microsoft Management Console (MMC)


SES, TAS, TIM


Application deployment


Centralized settings and replication


Windows Management Instrumentation (WMI)


All settings can be configured programmatically


Uniform access for mgmt data from multiple sources


Sharing mgmt data across multiple mgmt applications


App Provisioning Tool

47

Monitoring your MSS Deployment


Real
-
time monitoring facilities


System monitor (disk, memory, proc, net, etc)


Performance monitor (counters, traces, alerts)


Centralized Monitoring via MMC


SNMP


Events can be configured to generate traps


MOM Management Pack


Log Analysis Tools


48

Data Analysis Framework

Application

49

Reporting and Tuning


Scenarios


Business Manager


how many calls? which services? how many
successful transactions?


Application Developer


how good are my grammars? where are people
hanging up/transferring?


System administrator


where are the system latencies? what is causing
trace warning messages?


MSS solutions


Speech Application Reports


Call Viewer

50

Call Viewer


Analyst selects problematic calls


queries over times, events, particular
properties


quickly retrieves all calls symptomatic of the
problem


Drills down into the call


flexible views, according to event filters
selected


turn by turn


including user audio


component event by component event


detailed information


even finer
-
grained configuration


conditional event filtering

51

Call Viewer

52

Reports

Call
Aggregates

Provides total calls aggregated by various time periods

Call Volume

Displays call volume across a single span of time, plus providing
comparative data on disconnection types, transferred calls, and call
duration.

Dialog
Overview

For the selected time period, displays data on QA use and QA
recognition confidence scores, plus command use.

Errors

Displays summary data on call completion, a count of tasks missing
TaskStart and TaskComplete events, and occurrence counts for error
event types.

Messages

Provides counts of messages sent by the SALT LogMessage function,
and a breakdown by message name and content.

Prompts

Provides counts for bargein and user
-
perceived latencies, prompt
ending status, individual prompt totals, and comparison of text
-
to
-
speech versus recorded prompts.

Server
Statistics

Provides data on latencies, channel and port usage, and answering
time.

Turn Analysis

Displays data on QA usage per call, sums and percentages for QA
History values, compares speech versus DTMF input, and reports on
time spent on pages.

Tasks

Reports on task completion, duration and progress, and the count of
QAs used for each task.

53

© 2003
-
2004 Microsoft Corporation. All rights reserved.

This presentation is for informational purposes only. Microsoft makes no warranties, express or implied, in this summary.