Vocalizer 30 Developers Guide PDF

Nuance Vocalizer
3.0
Developers Guide

Nuance Vocalizer 3.0
Developers Guide
Copyright 2004 Nuance Communications, Inc. All rights reserved.
1005 Hamilton Avenue, Menlo Park, California 94025 U.S.A.
Printed in the United States of America.
Last updated July 2004.
Information in this document is subject to change without notice and does not represent a
commitment on the part of Nuance Communications, Inc. The software described in this document is
furnished under a license agreement or nondisclosure agreement. The software may be used or copied
only in accordance with the terms of the agreement. You may not copy, use, modify, or distribute the
software except as specifically allowed in the license or nondisclosure agreement. No part of this
document may be reproduced or transmitted in any form or by any means, electronic or mechanical,
including photocopying and recording, for any purpose, without the express written permission of
Nuance Communications, Inc.
Nuance Vocalizer uses the Regex++ library distributed by Boost.org. Copyright 1998-2001 Dr. John
Maddock.
Nuance and Nuance Communications are registered trademarks of Nuance Communications, Inc.
SpeechChannel, SpeechObjects, Nuance Verifier, and Nuance Voice Web Server are trademarks of
Nuance Communications, Inc. Any other trademarks belong to their respective owners.
Contents iii
Contents
About this guide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Audience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Related documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x
Typographical conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
Where to get help . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii
Chapter 1. Introducing Nuance Vocalizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Chapter 2. Installing Nuance Vocalizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Supported platforms and system requirements . . . . . . . . . . . . . . . . . . . . . . 3
Memory requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Third-party software requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Installing Nuance Vocalizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Installing Vocalizer on Windows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Installing Vocalizer on Solaris . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Environment variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Nuance Vocalizer licensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Testing your installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Uninstalling Vocalizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Installing voice packs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Installing voice packs on Windows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Installing voice packs on Solaris . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Chapter 3. Configuring Vocalizer with the Nuance System . . . . . . . . . . . . . . . . . 9
Including Vocalizer in a Nuance configuration . . . . . . . . . . . . . . . . . . . . . . . 9
Starting a Vocalizer TTS server . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Command-line options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Developers Guide
iv

Nuance parameter settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Connecting Vocalizer to a resource manager . . . . . . . . . . . . . . . . . . . . 17
Running Vocalizer without a resource manager . . . . . . . . . . . . . . . . . 19
Using the Nuance Watcher to start Vocalizer . . . . . . . . . . . . . . . . . . . . . . . 20
Starting Vocalizer automatically with a Watcher startup file . . . . . . 21
Communicating with Vocalizer through the Watcher . . . . . . . . . . . . 22
Enabling Vocalizer logs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Directing requests to a specific TTS server . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Chapter 4. Playing TTS prompts from your application . . . . . . . . . . . . . . . . . . . 27
Using Nuance prompt playback functions . . . . . . . . . . . . . . . . . . . . . . . . . 27
SpeechObjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
NuanceSpeechChannel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
RCEngine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Using TTS in a VoiceXML application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
VoiceXML 1.0 elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
SSML 1.0 elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
SSML sample text . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Resolving XML parsing errors with predefined entities . . . . . . . . . . . . . . 39
Encoding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
URI formats for the SSML <audio> element . . . . . . . . . . . . . . . . . . . . . . . . 40
Chapter 5. Starting Vocalizer through the Launcher . . . . . . . . . . . . . . . . . . . . . 43
Using the Vocalizer Launcher . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Configuration editor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Status window . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Chapter 6. Preprocessing text input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Passing email messages to Vocalizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Features of the Email Reader Configuration Tool . . . . . . . . . . . . . . . . 47
Using the Email Reader Configuration Tool . . . . . . . . . . . . . . . . . . . . 48
Distributing changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Creating custom filters for text replacement . . . . . . . . . . . . . . . . . . . . . . . . 52
Applying filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Using the Text Replacement Filter Editor . . . . . . . . . . . . . . . . . . . . . . . 53
Contents v
Distributing changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
Chapter 7. Generating audio files from text input . . . . . . . . . . . . . . . . . . . . . . . . 55
Using the Offline Audio Generator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Validating multi-server configurations . . . . . . . . . . . . . . . . . . . . . . . . . 57
Chapter 8. Customizing your applications dictionary . . . . . . . . . . . . . . . . . . . . . 59
Features of the Dictionary Editor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Audio feedback . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
Word details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Phoneme set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Audio scratchpad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Using the Dictionary Editor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Enabling logging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Adding new entries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Modifying dictionary entries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Testing your entries in Scratchpad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Deleting entries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Distributing dictionary changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
Chapter 9. Techniques for enhancing audio output . . . . . . . . . . . . . . . . . . . . . . 69
Phrasing and punctuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Periods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
Hyphens and dashes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
Parentheses and double quotes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
Apostrophes and single quotes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Single-letter abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Ambiguous abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Homographic abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Measurement and numeric abbreviations . . . . . . . . . . . . . . . . . . . . . . 75
Acronyms and initials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Note on capitalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Cardinals and ordinals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Developers Guide
vi

Digit sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Fractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Decimal fractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Percentages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Account and social security numbers . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Combining letters and numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Currency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Telephone numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Addresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Zip codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Dates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Email and web addresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Handling 8-bit characters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Appendix A. Text processing for Canadian French . . . . . . . . . . . . . . . . . . . . . . . 87
Text encoding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Phrasing and punctuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Periods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Hyphens and dashes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Parentheses and double quotes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Single-letter abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Homographic abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Measurement and numeric abbreviations . . . . . . . . . . . . . . . . . . . . . . 91
Acronyms and initials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
Capitalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Cardinals and ordinals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Digit sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Fractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Decimal fractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Contents vii
Percentages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Account numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Combining letters and numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
Currency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Telephone numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
Addresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
Dates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
Email and web addresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
Appendix B. Text processing for American Spanish . . . . . . . . . . . . . . . . . . . . . 103
Text encoding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Regionalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Phrasing and punctuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Periods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Parentheses and double quotes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
Single-letter abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
Homographic abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Measurement and numeric abbreviations . . . . . . . . . . . . . . . . . . . . . 107
Acronyms and initials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
Capitalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
Cardinals and decimal fractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
Ordinals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
Digit sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Fractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Percentages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Account numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Combining letters and numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Currency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Telephone numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
Dates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
Times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Developers Guide
viii

Addresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
Email and web addresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
About this guide
Audience
ix
About this guide
Nuance Vocalizer provides text-to-speech (TTS) services integrated with the
Nuance Systems distributed architecture.
This guide describes how to use Nuance Vocalizer to incorporate text-to-speech
in a Nuance speech application, including how to invoke TTS from your
application code and how to set up your runtime configuration to most
efficiently use Nuance Vocalizer services.
Audience
This document is for developers creating speech applications based on the
Nuance Speech Recognition System, including applications built with
Foundation SpeechObjects or applications built with VoiceXML running on
the Nuance Voice Web Server.
Organization
The guide is organized as follows:
Chapter 1 introduces Nuance Vocalizers features.
Chapter 2 covers the installation process and system requirements for running
Nuance Vocalizer.
Chapter 3 describes how to start and configure Nuance Vocalizer with the
Nuance System.
Chapter 4 provides information on invoking TTS from your application code and
describes Nuance Vocalizers support for speech synthesis markup tags.
Chapter 5 explains how to start and modify configuration settings for Nuance
Vocalizer using the Launcher, a graphical tool included with your installation.
Chapter 6 outlines the procedures for preprocessing text input for email reading
and creating text replacement filters.
Developers Guide
x

Chapter 7 explains how to generate audio files from text input through the
Offline Audio Generator tool.
Chapter 8 describes how to use the Dictionary Editor tool to customize the
pronunciation dictionary for your applications.
Chapter 9 describes how English text input is handled by the TTS engine.
Appendix A provides information on how Canadian French text input is
processed.
Appendix B provides information on how American Spanish text input is
processed.
Related documentation
The Nuance documentation contains a set of developer guides as well as
comprehensive online API reference documentation.
In addition to this guide, the documentation set includes:
Introduction to the Nuance System, which provides a comprehensive overview

of the Nuance System architecture and features, the available tools and
programming interfaces, and the speech application development process.
Nuance System Installation Guide, which describes how to install and

configure the Nuance Speech Recognition System.
Nuance Grammar Developers Guide, which describes the process of designing

grammars and creating a recognition package.
Nuance Application Developers Guide, which describes how to develop,

configure, and tune a Nuance speech application. This manual describes
general application development and deployment issues such as setting
Nuance parameters and launching recognition clients, servers, and other
required processes.
Nuance System Administrators Guide, which describes topics related to

monitoring the performance of a running speech application, including
controlling Nuance process log output, using the Nuance Watcher to manage
processes across a network, and using the runtime task adaptation feature.
Nuance Verifier Developers Guide, which describes how to use the Nuance
Verifier to add security features to speech recognition and IVR
applications.
Nuance Platform Integrators Guide, which describes the process and

requirements for integrating the Nuance System with an existing IVR
About this guide
Typographical conventions
xi
(interactive voice response) platform or toolkit. It provides detailed
information on the primary Nuance integration APIs, the RCEngine and
NuanceSpeechChannel, and also discusses other specialized APIs for adding
custom components such as audio providers to your integration.
Nuance System Glossary, which defines terms and acronyms used in the
Nuance documentation set.
Nuance API Reference, which includes HTML-based documentation for

Nuance APIs, parameters, command-line utilities, and GSL (grammar
specification language) syntax. To access this documentation, open the file
%NUANCE%\doc\api\index.html.
You may also refer to the Foundation SpeechObjects documentation, including
the SpeechObjects Developers Guide, which describes the Nuance SpeechObjects
framework and how to use it to build a speech recognition application.
SpeechObjects documentation is shipped with the Foundation SpeechObjects
product.
To view the entire set of documentation online, open the file
%NUANCE%\doc\index.html in any HTML browser. Install the
%NUANCE%\doc directory on an internal web server or file system to
minimize space requirements. This directory includes all Nuance developer
guides in both HTML and PDF format, and HTML reference documentation.
Typographical conventions
Nuance manuals use the following text conventions:
italic text Indicates variables, file and path names, program names,
program options, web and email addresses, as well as
terms introduced for the first time. For example:
Edit the ag.cfg configuration file.
Courier New Indicates method/member functions, parameter names
and values, program constants, and onscreen program
output. For example:
Nuance recommends that you set the parameter
audio.OutputVolume to 255.
Developers Guide
xii

Note: The Nuance System runs on both Windows and Unix platforms. Windows
syntax is typically used throughout the documentation. If you are running on a
Unix platform, substitute Unix syntax. For example, use $NUANCE wherever
%NUANCE% appears, and use / in place of \ in path names. Differences in
usage or functionality on Windows and Unix platforms are noted where
relevant.
Where to get help
If you have questions or problems, Nuance provides technical support through
the Nuance Technical Support Online (Nuance Technical Support Online), a
web-based resource center that includes online forums, technical support guides
on specific topics, access to software updates, as well as a support request form.
If you are a member, go to extranet.nuance.com to log on, or see the Nuance
website www.nuance.com for information on how to become a member.
To submit comments on the documentation, please send email directly to
[email protected]. Note that no technical support is provided through this
email address. Technical support is provided through the Nuance Technical
Support Online.
> Courier New Indicates commands or characters you type in and the
responses that appear on the screen. The > character
indicates the MS-DOS command prompt or Unix shell.
Everything after this character is intended to be typed in.
For example:
> resource-manager
In this example, the text that you actually type at the
command line is resource-manager
Courier New Indicates a value that you replace. For example:
The usage for the nlm utility is:
> nlm license_key
In this example, license_key is a value that you replace,
so the text you actually type could be, for example:
> nlm ncr8-16-100-a-b22-333c4d55eeff
Chapter 1 Introducing Nuance Vocalizer
Overview
1
1
Chapter 1
Introducing Nuance
Vocalizer
Nuance Vocalizer is Nuances best-of-breed, text-to-speech (TTS) server,
enabling automated generation of high quality speech synthesis with natural
voice quality. Nuance Vocalizer integrates seamlessly with the Nuance Speech
Recognition System and Nuance Verifier for a complete speech solution.
This chapter introduces Vocalizers features, the types of applications that can
implement TTS functionality, and the benefits of using TTS.
Overview
Nuance Vocalizer is designed to be used either as a standalone TTS server or as
part of a single-tier or distributed architecture.
Vocalizer supports many features that allow you to customize speech output
and enhance performance, including:
Support for Windows 2000 and SPARC Solaris platforms
Compliance to VoiceXML 1.0 and the World Wide Web Consortiums (W3C)
April 5, 2002 Working Draft for Speech Synthesis Markup Language
Specification
Dynamic editing of entries and pronunciations in user dictionaries
Customized text preprocessing for email reading and text replacement filters
Generation of audio files in various formats from text files or direct user
input
Ability to run voice data from disk, thus reducing system memory
requirements
Support for these voice packs:

Developers Guide
2

The voice names correspond to the Nuance standard persona names. For
information on selecting and running voices, see Selecting voices on page
12.
Caution: When using the non-English voice packs and entering text in SSML
mode, you must specify the encoding type (ISO-8859-1) to avoid unpredictable
behavior. See Encoding on page 39.
Voice pack Language Gender
Laurie Woods North American English Female
Reed Johnston North American English Male
Claire Kingston U.K. English Female
Tim Cooper U.K. English Male
Sarah Brown Australian/New Zealand Female
Josh Donnelly Australian/New Zealand Male
Julie Deschamps Canadian French Female
Catalina Romero American Spanish Female
Chapter 2 Installing Nuance Vocalizer
Supported platforms and system requirements
2
3
Chapter 2
Installing Nuance Vocalizer
This chapter defines system requirements supported by Nuance Vocalizer and
describes the procedure for installing Vocalizer and supported voices.
Supported platforms and system requirements
Nuance Vocalizer runs on both Windows 2000 and SPARC Solaris 2.8 platforms
and can be installed either:
With the Nuance System

Nuance Vocalizer works with all audio/telephony systems supported on the
underlying version of the Nuance System. For installation information, see
the Nuance System Installation Guide.
Standalone (without the Nuance System installed)

You can install and run Vocalizer independently of the Nuance System.
However, Vocalizer uses the Nuance License Manager (NLM). If you plan to
run the nlm process from another machine where the Nuance System is
installed, make sure you also have the appropriate Nuance System service
pack installed on that machine.
See the Release Notes for specific requirements.
Developers Guide
4

Memory requirements
Nuance recommends the following minimum memory requirements:
For 1-20 channels, 512 MB
For 20-50 channels, 768 MB
For 50 channels, 1 GB
Note: These recommendations are approximate figures, as actual memory
requirements depend on the maximum prompt duration used.
Third-party software requirements
The GUI utilities included with Nuance Vocalizer require the Java 2 Runtime
Environment (JRE) version 1.3 (or higher).
On Windows, the JRE version 1.3 is included as part of the installation

package. See Installing Vocalizer on Windows on page 5.
If you upgrade to a newer version of the JRE after installing Vocalizer and the
GUI utilities, you must update the path pointing to the JRE within the .bat
files for each of the utilities:
a From %VOCALIZER%\bin\win32 (on Windows), open each of these files
with a text editor such as Notepad:
DictionaryEditor.bat
EmailReaderConfigTool.bat
OfflineAudioGenerator.bat
VocalizerLauncher.bat
ReplacementFilterEditor.bat
b In each of the files, change the path to the javaw executable. For example,
if your path is:
"C:\Program Files\JavaSoft\JRE\1.3\bin\javaw"
change this to the location of the latest version, for example:
"C:\Program Files\JavaSoft\JRE\1.4\bin\javaw"
On Solaris, you must download and install the JRE if you do not have it
already. See the Sun website (www.sun.com) for download information.
5
The Nuance Vocalizer installation package is available for download from the
Nuance Technical Support Online website (support.nuance.com). After
downloading the appropriate installation package, follow the installation
procedures described next.
Installing Vocalizer on Windows
To install Vocalizer on Windows:
1 Double-click the downloaded executable file to run InstallShield.
a The opening dialogue window offers the option of including these GUI
tools and the JRE version 1.3 in your installation:
Dictionary Editor
Email Reader Configuration Tool
Offline Audio Generator
Text Replacement Filter Editor
Vocalizer Launcher
Java 2 Runtime Environment (JRE) version 1.3

In a multi-server configuration, you do not need to install these tools on
every machine, since modifications made using these tools can be
distributed to other machines. Before including them in your installation,
consider:
If you do not plan to use the GUI tools on this machine, you are not
required to install them or the JRE
If you already have the JRE version 1.3 (or higher) installed, do not
include the JRE in your installation
If you have an earlier version of the JRE, uninstall it before installing

version 1.3
See Third-party software requirements on page 4.
b You also have the option of changing your installation path. By default,
Vocalizer is installed in the Program Files\Nuance\Vocalizer3.0 directory
and the %VOCALIZER% environment variable is set accordingly.
Vocalizer executables are installed in the %VOCALIZER%\bin\win32
directory.
Developers Guide
6

2 Reboot your machine once installation is complete.
Installing Vocalizer on Solaris
To install Vocalizer on SPARC Solaris:
1 Log on as the root user
2 Copy the downloaded installation package into a temporary directory, for
example
3 Gunzip the file:
gunzip installation-package.tar.gz
4 Untar the file:
tar xvf installation-package.tar.gz
5 Run the install.sh script and follow the instructions:
./install
6 Set the $VOCALIZER environment variable to the installation path where
Vocalizer was installed, with either bash or tcsh:
bash-$] export VOCALIZER=Vocalizer_path
tcsh-$] setenv VOCALIZER Vocalizer_path
Nuance recommends that this environment variable be set in a persistent
place, such as in a .profile (bash) or .cshrc (tcsh) file.
7 Set the $NUANCE environment variable:
bash-$] export NUANCE=$VOCALIZER/core
tcsh-$] setenv NUANCE $VOCALIZER/core
Note: $NUANCE may be set to $VOCALIZER/core or to an existing Nuance
System installation.
8 Source the SETUP scripts in $NUANCE and $VOCALIZER/bin/sparc-solaris to
set up your PATH and LD_LIBRARY environment variables, with either bash
or tcsh:
With bash
bash-$] cd $NUANCE
bash-$] . SETUP sh
bash-$] cd $VOCALIZER/bin/sparc-solaris
bash-$] . SETUP sh
With tcsh
Testing your installation
7
tcsh-$] cd $NUANCE
tcsh-$] source SETUP
tcsh-$] cd $VOCALIZER/bin/sparc-solaris
tcsh-$] source SETUP
Nuance recommends that this environment variable be set in a persistent
place, such as in a .profile (bash) or .cshrc (tcsh) file.
Environment variables
On Windows, the Vocalizer installation process modifies your systems PATH
variable to include necessary libraries. Make sure the PATH variable includes
%NUANCE%\bin\win32 and %VOCALIZER%\bin\win32.
On Unix, $VOCALIZER/bin/sparc-solaris is added to the PATH and LD_LIBRARY
environment variables when you source the SETUP script, as described in the
installation procedure for Solaris (step 8 on page 6).
Nuance Vocalizer licensing
Nuance provides a license key that controls the number of concurrent channels
that can be opened to the TTS server. You can run this license file using an
instance of the Nuance License Manager (nlm), as described in Starting a
Vocalizer TTS server on page 11.
Note: The default two-port license shipped with Vocalizer uses port 8471. When
you buy your Vocalizer license, this port number may change, and you must set
the lm.Addresses parameter accordingly.
Testing your installation
You can test your installation by sending text to Vocalizer to be synthesized
through the Offline Audio Generator, a graphical tool included with your
installation. See Chapter 7 for information on using this tool.
Uninstalling Vocalizer
When you uninstall Vocalizer, all user data, including custom dictionary files,
email configuration files, voice configuration files, and custom filters, is backed
up into a temporary folder, vocalizer_files-<date>-<time>, in either the %TEMP%
or %SYSTEMROOT% directory.
Developers Guide
8

Installing voice packs
The voice packs are available for download from the Nuance Technical Support
Online website (support.nuance.com) and do not require separate licensing.
Vocalizer servers only support one voice at a time. When you start your
Vocalizer server, you can select which voice to use through the -voice
command-line option. See Selecting voices on page 12.
Installing voice packs on Windows
To install a voice pack on Windows:
1 Double-click the voice pack executable file to run InstallShield.
2 You will be prompted for the installation path.
By default, InstallShield detects your Vocalizer installation and installs the
voice pack files in the appropriate directory.
Installing voice packs on Solaris
To install a voice pack on Solaris:
1 Log on as the root user.
2 Copy the voice pack file into a temporary directory:
cp VoiceName_pack.tar.gz tmp
3 From that directory, uncompress the file:
gunzip VoiceName_pack.tar.gz
4 Untar the file:
tar xvf VoiceName_pack.tar
5 Run the install script:
./install-VoiceName.sh
The script detects your Vocalizer installation and installs the voice pack files
in the appropriate directory.
Chapter 3 Configuring Vocalizer with the Nuance System
Including Vocalizer in a Nuance configuration
3
9
Chapter 3
Configuring Vocalizer with
the Nuance System
This chapter describes how to start the processes that let you use Nuance
Vocalizer text-to-speech services from a Nuance speech application. You can
start these processes directly from the command line, or via the Nuance Watcher
tool.
Note: This chapter describes the runtime configuration for supporting
text-to-speech. For information on Nuance programming interfaces for
incorporating text-to-speech in your application code, see Chapter 4.
Vocalizer runs as an integrated part of a Nuance configuration, using
functionality provided by the recognition client to issue TTS requests to
Vocalizer and play the synthesized speech back to the caller.
The following diagram shows the general flow of data between the application
and the TTS engine:
Developers Guide
10

Vocalizer runs as a server process similar to other Nuance processes such as the
recognition and compilation servers. You can set up a runtime configuration
including Vocalizer TTS servers in one of two ways:
Connect each recognition client directly to a single Vocalizer TTS server

process
Use a Nuance resource manager to mediate requests between recognition

clients and one or more Vocalizer TTS servers
Nuance recommends that you manage Vocalizer server connections through a
resource manager. The resource manager enables load balancing and scalability
for large configurations. It also allows you to add, remove, and restart TTS
server processes without restarting the entire configuration.
The following sections describe how to start a Vocalizer TTS server and how to
integrate it within runtime configurations both with and without a resource
manager.
Note: If you are not familiar with the resource manager and the overall Nuance
System runtime architecture, see the Introduction to the Nuance System guide
shipped with the Nuance System.
11
Starting a Vocalizer TTS server
To start a Vocalizer TTS server, you need to run nlm before running the vocalizer
command-line program:
1 Start the license manager in a command-prompt window. From the
%VOCALIZER% directory, enter:
> nlm .\license.txt
2 Start vocalizer in a separate command-prompt window. From the
%VOCALIZER%\bin\win32 directory, include the hostname and port of the
machine running the nlm process, setting the port to 8471. For example:
> vocalizer lm.Addresses=licenseServer:8471
Note: The default two-port license shipped with Vocalizer uses port 8471.
When you buy your Vocalizer license, this port number may change, and
you must set the lm.Addresses parameter accordingly.
Command-line options
This section describes command-line options you can use when starting the
vocalizer program. If you run vocalizer without including any options, default
settings for speech synthesis are applied.
Specifying the
number of channels
You can use the --num_channels option to specify the maximum number of
simultaneous requests Vocalizer will handle. Requests exceeding the specified
maximum experience higher latency (time to first audio) than other requests.
For example, if you set -num_channels 10, and Vocalizer is already handling nine
requests, the tenth request is handled immediately, but subsequent requests are
blocked until a TTS request completes and Vocalizer can handle a new request.
Memory issues Vocalizers memory requirements are dependent upon the number of channels
specified with the -num_channels option, so make sure you do not specify more
channels than you need so that memory is not wasted. Even a relatively small
number of channels (10-20) could affect performance on machines with small
amounts of memory (512 MB) due to memory swapping.
License issues Vocalizer allows you the option of having all licenses allocated at startup or
having them allocated as they are required. By default, licenses are granted on a
per request basis.
If you choose to preallocate your licenses, make sure -num_channels does not
exceed the number of ports available with your license. When you start
Vocalizer, it attempts to grab the specified number of licenses from the license
server. If there are not enough licenses, Vocalizer will not start.
Developers Guide
12

To enable preallocation of licenses, include the option -preallocate_licenses on the
command line.
Selecting voices By default, Vocalizer uses the North American English female voice, Laurie
Woods.
To select a different voice when starting vocalizer, use the -voice option. For
example, if you want to use the Reed Johnston voice pack, enter:
> vocalizer lm.Addresses=hostname:8471 -voice ReedJohnston
Vocalizer currently supports these voice packs:
You can download these voice packs from the Nuance Technical Support Online
website (support.nuance.com). See Installing voice packs on page 8.
Note: See Appendix A for information on Vocalizers handling of Canadian
French text input, and Appendix B for American Spanish information.
Using the -region
option
For American Spanish, Vocalizer supports various conventions for numbers,
dates, and addresses. To indicate how Vocalizer should interpret Spanish text
input, set the -region option to either US or CALA. For example:
vocalizer lm.Addresses=hostname:8471 -voice catalinaromero
region CALA
For information on the differences between these settings, see Regionalism on
page 104.
Running voice data
from disk
Vocalizer also allows you to run the voice pack from a disk in order to reduce
memory requirements by about 250 MB at startup.
To enable this feature, include the -voices_from_disk option. For example:
Language Gender Voice pack Command-line setting
North American English Female Laurie Woods lauriewoods
North American English Male Reed Johnston reedjohnston
U.K English Female Claire Kingston clairekingston
U.K English Male Tim Cooper timcooper
Australian English Female Sarah Brown sarahbrown
Australian English Male Josh Donnelly joshdonnelly
Canadian French Female Julie Deschamps juliedeschamps
American Spanish Female Catalina Romero catalinaromero
13
vocalizer lm.Addresses=hostname:8471 -voice reedjohnston
-voices_from_disk
Note: At least one voice pack must be installed on your system before using this
option.
Using the -pruning
option
The -pruning option provides another means of controlling CPU usage. The
-pruning option can be set to a value between 80 to 330. Default pruning values
are specific to voice packs.
Modifying the -pruning value enables you to increase or decrease the number of
channels that can be obtained for the same CPU type or speed, such as on a dual
1.4 GHz Pentium III. Lower -pruning levels (towards 80) allow for more
channels, but may produce slightly degraded output. If you require extremely
high quality output, you can increase the -pruning level, which reduces the
number of channels supported. Changing the value for -pruning has little impact
on overall memory consumption.
Selecting the audio
encoding
Vocalizer can generate speech output encoded in the following formats:
To select an encoding, use the -encoding option:
> vocalizer lm.Addresses=hostname:8471 -encoding format
where format is either alaw or mulaw.
Note: Make sure that your audio provider supports the encoding that you
specify.
Text format options Vocalizer can be configured to automatically determine what type of text is
being sent to it and treat it appropriately. To do so, you must give Vocalizer a
hint as to what text types are expected by:
Specifying the markup language, either plain text, VoiceXML, or SSML
Setting up Vocalizer to accept email files

Setting the default
markup type
Vocalizer lets you specify the text to be synthesized using either plain text,
VoiceXML markup tags, or the W3Cs Speech Synthesis Markup language
(SSML) tags. By default, Vocalizer uses plain text.
Format Usage
mulaw Mulaw encoding used as a standard in North America and
Japan. This is the default.
alaw Alaw encoding used as a standard in Europe.
Developers Guide
14

Use the -text_type option to specify which format you are using. For example, if
you are using SSML tags, start vocalizer with the following setting:
> vocalizer lm.Addresses=hostname:8471 -text_type ssml
See Using TTS in a VoiceXML application on page 29 for information on how
markup tags are supported.
Enabling the email
parser
Vocalizer can parse and read email messages encoded in the RFC 822 standard.
To enable this feature, specify the -filter option on the vocalizer command line:
> vocalizer lm.Addresses=hostname:8471 -filter email
See Passing email messages to Vocalizer on page 47 for information on how
you can configure the email parser.
Using custom filters You can also use the -filter option to specify multiple filters, using a
comma-separated list. (Do not include spaces between items.) For example,
>vocalizer -filter filter1,filter2,[other_filters]
See Creating custom filters for text replacement on page 52.
Expected text formats Setting the -text_type and -filter options enables Vocalizer to expect the correct
data format and treat it appropriately, as follows:
For more information on creating custom filters, see Creating custom filters for
text replacement on page 52.
Option summary The full usage statement for the vocalizer command-line tool is:
> vocalizer lm.Addresses=hostname:8471
[-num_channels number]
[-preallocate_licenses]
Command-line option Expected format
-text_type ssml Plain text or SSML
-text_type vxml Plain text or VoiceXML
-text_type plaintext Plain text
-text_type ssml -filter email Plain text, SSML, or email
-text_type vxml -filter email Plain text, VoiceXML, or email
-text_type plain -filter email Plain text or email
-filter email Plain text or email
15
[-text_type type]
[-voice voice_name]
[-region region]
[-voices_from_disk]
[-pruning value]
[-encoding format]
[-filter email]
[-dictionary_port port_number]
[-help]
This table summarizes the settings for each option:
Option Description Type Values Default
-num_channels Specifies the number of engines this
instance of Vocalizer will support. See
Specifying the number of channels on
page 11.
Integer number 2
-preallocate_
licenses
Indicates that licenses are to be
preallocated at startup. By default, this
option is not included and licenses are
granted per request as described in
License issues on page 11.
String n/a n/a
-text_type Specifies the markup language used in the
text to be passed to the Vocalizer TTS
engine. See Selecting the audio
encoding on page 13.
String plain
ssml
vxml
plain
-voice Specifies the voice to use for synthesis.
See Selecting voices on page 12.
String lauriewoods
reedjohnston
clairekingston
timcooper
sarahbrown
joshdonnelly
juliedeschamps
catalinaromero
LaurieWoods
-region Specifies the convention to use when
interpreting number, date, and address
text input for American Spanish. See
Regionalism on page 104.
String US
CALA
US
-voices_from_
disk
Specifies to run the voice pack from disk.
See Running voice data from disk on
page 12.
n/a n/a n/a
Developers Guide
16

Nuance parameter settings
The vocalizer program can also accept a number of Nuance configuration
parameters on the command line. These parameters control how the vocalizer
process interacts with other Nuance processes, and include:
rm.Addresses
Defines the location (machine and port) of one or more resource managers
for the TTS server to connect to. See Connecting Vocalizer to a resource
manager on page 17 for more information.
-pruning Specifies the pruning level for the trade-off
between output quality and CPU usage.
Integer Any value between
80 and 330
inclusive
Specific to
voice pack
-encoding Selects the audio encoding to be used to
generate the speech output.
String mulaw
alaw
mulaw
-filter Indicates how text input will be processed
by the TTS engine. You can specify
multiple filters with this option, including:
emailTo enable preprocessing of

email messages for appropriate
synthesis. See Enabling the email
parser on page 14.
Custom filtersTo enable text

replacement. See Using custom filters
on page 14.
For example,
>vocalizer -filter email,
filter1,filter2
String email
custom_filters
No default
-dictionary_port Specifies the port for phoneme translation
with the Dictionary Editor tool. See
Chapter 8.
String port_number 22552
-help
-h
Displays the usage statement for the
vocalizer program.
When you include this option on the
command line, vocalizer displays the
usage statement and exits, regardless of
other options you specify.
n/a n/a n/a
Option Description Type Values Default
17
lm.Addresses
Defines the location (machine and port) of the machine running the Nuance
License Manager (nlm) process for Vocalizer. Set the port to 8471. See
Starting a Vocalizer TTS server on page 11.
tts.Port
Specifies the port the TTS server uses to listen for requests. The default is
32323. You only need to set this if you want to use a specific port number
other than the default.
If you are running two or more Vocalizer TTS servers on the same machine
you must set this for all but one TTS server process. The first instance you
start uses the default port, so you must specify a different port for additional
servers.
tts.ResourceName
Provides a name for an instance of the Vocalizer TTS server. Clients can use
this name to direct a TTS request to a specific class of servers. See Directing
requests to a specific TTS server on page 24.
config.LogFileRootDir and related parameters
Set this parameter to cause the TTS server process to output log files. Other
config.LogFile* parameters control how log files are serially numbered
and how many files are produced. If you start vocalizer via the Nuance
Watcher, you should enable this logging, as the program output is lost
otherwise. See Enabling Vocalizer logs on page 23 for more information.
Connecting Vocalizer to a resource manager
Nuance recommends running Vocalizer in a configuration in which the Nuance
resource manager controls the recognition clients access to the TTS server. This
allows multiple recognition clientseach of which can support multiple ports
to use Vocalizers TTS services through one or more Vocalizer TTS server
processes. Whenever a recognition client requests a text-to-speech operation, the
resource manager identifies the least-busy TTS server, and the client connects to
that server for the duration of the operation. Nuance recommends this
configuration even if you are only running a single TTS server, as it allows you
to transparently add or swap out additional servers to the recognition clients.
A small-scale configuration of this type might look like this:
Developers Guide
18

To start this type of configuration:
1 Start the resource-manager process:
> resource-manager
2 Start nlm with the Vocalizer license:
> nlm .\license.txt
3 Start the vocalizer process, specifying the location of the resource manager
with the rm.Addresses parameter. For example, if you started the resource
manager on a machine named roadrunner:
> vocalizer rm.Addresses=roadrunner
lm.Addresses=hostname:8471
If your configuration includes multiple resource managers on different
machines, you can set rm.Addresses to a comma-separated list of all
machines running resource managers, for example:
> vocalizer rm.Addresses=roadrunner,coyote
If you run the resource manager on a specific port (other than the default),
include the port number in the rm.Addresses setting, for example:
Recognition
server
Resource
manager
Recognition
client
Nuance speech application Nuance speech application
Recognition
server
Vocalizer
Recognition
client
19
> resource-manager rm.Port=8584
> vocalizer rm.Addresses=roadrunner:8584
This causes the vocalizer process to register with the resource manager, and
any application requests for TTS generation are directed to the Vocalizer
server.
Note: See the Nuance System documentation for information on starting the
other processes in your configuration, including the recserver. You can start
vocalizer at any point after the resource managers have been started (that is,
before or after the recognition servers and other processes).
Running Vocalizer without a resource manager
You can also run Vocalizer without a resource manager. In this case the
recognition client sends all TTS requests directly to a specific Vocalizer TTS
server. This type of configuration looks like this:
To start this type of configuration:
1 Start nlm with the Vocalizer license:
> nlm .\license.txt
2 Start the vocalizer TTS server process before starting your application or
recognition client:
> vocalizer lm.Addresses=hostname:8471
Recognition
client
Nuance speech application Nuance speech application
Vocalizer
Recognition
client
Developers Guide
20

3 Use the Nuance parameter client.TTSAddresses to specify the location of
the running TTS server when you start your application. For example, if you
have an application you start with the command-line program my_app that
accepts Nuance parameters on the command line:
> my_app client.TTSAddresses=coyote
Changing the port
setting
By default, Vocalizer uses the port 32323 to listen for direct requests from a
recognition client. To use a different port:
1 Set the tts.Port parameter on the vocalizer command line to the new port:
> vocalizer lm.Addresses=hostname:8471 tts.Port=5353
2 Include the port number in the client.TTSAddresses setting:
> my_app client.TTSAddresses=coyote:5353
If you are running multiple TTS servers, you can specify them all as a
comma-separated list in the client.TTSAddresses parameter, for example:
> my_app client.TTSAddresses=coyote,roadrunner
When the recognition client needs to issue a TTS request, it tries to connect to
the first TTS server in the list, then the second, and so on, until it successfully
connects to an available server.
Using the Nuance Watcher to start Vocalizer
The Nuance Watcher, shipped with the Nuance System, is an operations tool
that monitors Nuance processes and can restart them if they fail. You can also
use the Nuance Watcher to manually start and stop processes, and query the
parameter information associated with a given process. Vocalizer is integrated
with the Nuance Watcher, meaning that you can start and stop Vocalizer
through the Watcher, and have the Watcher restart the Vocalizer TTS server
process if it fails.
To use the Watcher, you run the Nuance watcher-daemon on each host machine in
your Nuance configuration. That Watcher can detect and communicate with any
Nuance processincluding resource managers, recognition servers, and TTS
serversrunning on the machine. Each Watcher can also communicate with
other Watchers on other machines in your network. Typically, you use one
master Watcher to monitor all Nuance processes on all machines in your
configuration:
Using the Nuance Watcher to start Vocalizer
21
Note: For complete information on the Nuance Watcher, see the Nuance
Application Developers Guide.
The following sections describe how to start Vocalizer through a Watcher
startup file, and how to communicate with Vocalizer through a running
Watcher.
Note: Nuance also recommends that you use the Nuance logging parameters
when you start processes via the Watcher, otherwise the log output is lost. See
Enabling Vocalizer logs on page 23 for information.
Starting Vocalizer automatically with a Watcher startup file
With the Watcher, you can create a startup file containing all the processes you
want to start on a given machine. Each line of your startup file contains the
command line for a Nuance process you want to start. You then pass this file to
the watcher-daemon using the watcher.DaemonStartupFilename, and the
Watcher attempts to start each process with the commands you specify.
To start Vocalizer through a Watcher startup file:
1 In your startup file, specify the process name vocalizer. You can pass in any
options you can pass to the vocalizer command-line executable, for example:
vocalizer -num_channels 15 lm.Addresses=hostname:8471
host_1
Watcher
Recognition
client
Watcher
host_4
Watcher Recognition
server
host_2
Resource
manager
Watcher
Vocalizer
host_3
Developers Guide
22

You typically also specify:
The watcher.RestartOnFailure parameter (this tells the Watcher to

restart the process if it dies)
Parameters to set up program logging

The following shows a simple startup file that starts a resource manager and
a TTS server. Use forward slashes in path specifications (all platforms) and
use a backslash to continue a command line on multiple lines. Make sure
there are no spaces before or after \.
# Start a resource manager
resource-manager lm.Addresses=acme\
watcher.RestartOnFailure=TRUE\
config.LogFileRootDir=d:/logs/rm/rmlog
config.LogFileMaxNum=20
# Start the Nuance Vocalizer TTS server
vocalizer -num_channels 15
watcher.RestartOnFailure=TRUE\
config.LogFileRootDir=d:/logs/tts/ttslog
config.LogFileMaxNum=20\
rm.Addresses=hostname lm.Addresses=acme:8471
2 On the watcher-daemon command line, set
watcher.DaemonStartupFilename to the name of your startup file:
> watcher-daemon watcher.DaemonStartupFilename=
c:\scripts\watcher_startup.txt
Communicating with Vocalizer through the Watcher
You can also communicate with Vocalizer through a running Watcher process,
including:
Starting a Vocalizer TTS server
Quiescing or killing a Vocalizer TTS server
Getting general runtime information on the Vocalizer TTS server such as the
process ID or the time the process was started
To start a Vocalizer TTS server through the Watcher Telnet interface, you could
send the following command:
> startvp vocalizer watcher.RestartOnFailure=TRUE \
config.config.LogFileMaxNum=d:/data/logs/tts
The Watcher lists Vocalizer TTS server processes as vocalizer and process
parameters.
Enabling Vocalizer logs
23
Note: For more information on information available through the Watcher, see
the Nuance Application Developers Guide.
Enabling Vocalizer logs
The Vocalizer TTS server process can generate logs similar to other Nuance
runtime processes. If you are using the Watcher, you should enable logging, as
standard program output is otherwise lost.
To enable logging:
1 Create a folder to store the log files.
2 On the vocalizer command line, set the parameter config.LogFileRootDir
to the location at which you want log files to be written. This enables logging
and sets both the location to which the files are written and the root name
for the files. For example:
> vocalizer config.LogFileRootDir=d:/data/logs/vocalizer/tts
causes Vocalizer to write log files named tts001, tts002, and so on to the
directory d:\data\logs\vocalizer.
Note: To set this parameter in a Watcher startup file, use forward slashes in
your path name instead of back slashes. On a Windows DOS command line,
you can use either format.
You can further control the behavior of the logging mechanism by setting these
parameters on the vocalizer command line:
config.LogFileMaxNum controls the number of sequential log files created.

If you set this to 5, for example, the logger creates tts001 through tts005 and
then restarts with tts001. The last version of the tts001 log file is replaced.
config.LogFileMaxSize sets the maximum size of each log file, in

megabytes. If you set this to 1, for example, the logger writes to tts001 until it
reaches 1 MB, and then closes tts001 and starts tts002.
Note: config.LogFileRootDir and related parameters must be set on the
command line for each process you want to generate logs, including vocalizer
and watcher-daemon. You cannot set these parameters globally, for example, in a
nuance-resources.site file.
Developers Guide
24

Directing requests to a specific TTS server
Vocalizer provides a mechanism that lets you direct a TTS request to a specific
Vocalizer TTS server at runtime. This is useful if you want to run multiple TTS
servers with different characteristics (for example, supporting different markup
tags), and select the server to use for each individual request.
To do this, you give each server a name via the tts.ResourceName parameter as
you start it, and use this name at runtime to indicate the server you want to use
for each request:
Note: You must include a resource manager in your configuration to be able to
hot swap TTS engines via the tts.ResourceName parameter at runtime.
To set up the configuration shown:
1 Start your resource manager:
> resource-manager
2 Start the Vocalizer TTS servers, providing a name for each by setting the
tts.ResourceName parameter on the vocalizer command line, for example:
> vocalizer lm.Addresses=hostname:8471 -text_type vxml
rm.Addresses=coyote
tts.ResourceName=VXMLserver
Resource
Manager
Recognition
Client
Nuance speech application
Vocalizer
"VXMLserver"
Vocalizer
"SSMLserver"
Sets tts.ResourceName
to "VXMLserver" or "SSMLserver"
Directing requests to a specific TTS server
25
> vocalizer lm.Addresses=hostname:8471 -text_type ssml
rm.Addresses=coyote tts.ResourceName=SSMLserver
tts.Port=32324 -dictionary_port 22553
Note: By default, Vocalizer opens two ports at startup, the TTS port (32323)
and the dictionary port for phoneme translation (22552). If you are running
more than one instance of Vocalizer on the same machine, only one instance
can use the default port settings. Make sure to modify these port settings for
other instances by setting the tts.Port and -dictionary_port on the
vocalizer command line.
3 From the client side, use the parameter tts.ResourceName to select a
specific TTS server.
If this parameter is set by the client, any TTS request is directed to a TTS
resource with the same name. If it is not set, it is simply directed to the first TTS
server identified by the resource manager. By changing the setting of the
parameter before issuing a request, you can send the request to any named TTS
resource running in your configuration.
Every Nuance API includes a mechanism for setting parameters at runtime.
When programming with the NuanceSpeechChannel, for example, use the
setStringParameter method:
sc.setStringParameter("tts.ResourceName", "SSMLserver");
CorePromptPlayer prompter = sc.getPromptPlayer();
prompter.appendTTS("Hello World");
prompter.play(false);
You can also set the parameter when you initialize your application. If most of
your prompts include VoiceXML markup tags, for example, you can set the
VoiceXML TTS resource as the default when you start your application. For
example, if you start your application with the command-line executable myapp,
you could run:
> myapp rm.Addresses=coyote tts.ResourceName=VXMLserver
All TTS requests go to the VXMLserver TTS server until tts.ResourceName is
explicitly reset by your application.
Note: To set Nuance parameters from your application command line, your
applications initialization code must pass command-line arguments through to
a NuanceConfig object. The mechanism for doing this depends on which
Nuance API you use to write your application.
You can start multiple Vocalizer TTS servers with the same name. If you need to
run multiple servers with the same characteristics to handle the volume of TTS
requests, start each with the same tts.ResourceName value. When the
recognition clients tts.ResourceName parameter is set to that name and a TTS
Developers Guide
26

request is issued, the resource manager locates the least-busy of the servers with
that name.
Chapter 4 Playing TTS prompts from your application
Using Nuance prompt playback functions
4
27
Chapter 4
Playing TTS prompts from
your application
Vocalizers TTS engine puts a human voice to your application interface,
producing intelligible, synthesized speech from text input. This chapter
describes how to:
Invoke a TTS prompt from applications written with Nuance development

APIs, including SpeechObjects, NuanceSpeechChannel, and the RCEngine
Specify a TTS prompt using VoiceXML

Note: This chapter describes the APIs for specifying TTS prompts within your
application code. At runtime, you must have a TTS server running and
configured as described in Chapter 3 for the TTS API calls to work.
For information on the conventions used to synthesize the text based on the
way you transcribe it, see Chapter 9.
Using Nuance prompt playback functions
All Nuance development APIs include mechanisms for specifying TTS prompts.
These mechanisms are similar to the way you work with prerecorded prompts,
except that you specify the text to synthesize instead of the name of an audio
file. You specify the text directly in your function call, for example, in the
RCEngine:PlayPrompts call.
The following sections describe the mechanisms in each Nuance API, including:
SpeechObjects
NuanceSpeechChannel
RCEngine
Developers Guide
28

SpeechObjects
The Foundation SpeechObjects, Nuances Java-based speech application
development framework, defines classes that let you create specific TTS prompt
objects from a simple text string. This class, TTSPrompt, implements the Playable
interface, which allows it to be appended to a prompt queue and played
directly:
To use the TTSPrompt class, just set the text string you want to be synthesized.
You can then append the prompt to the prompt queue of your
SpeechChannel. For example:
TTSPrompt tts = new TTSPrompt("Hello world");
tts.appendTo(sc);
Note: In this example, sc is the SpeechChannel object being used by your
application.
NuanceSpeechChannel
If you are programming with the NuanceSpeechChannel API directly instead of
using the Foundation SpeechObjects, use the appendTTS method defined in the
CorePromptPlayer interface to create TTS prompts. This method creates the
prompt and appends it to the prompt queue.
The NuanceSpeechChannel class includes a default implementation of the
CorePromptPlayer interface. For example, if sc is your NuanceSpeechChannel
object, the following appends and plays a TTS prompt:
CorePromptPlayer player = sc.getPromptPlayer();
player.appendTTS("Hello world");
player.play(false);
For more information about the NuanceSpeechChannel API, see the online
documentation shipped with the Nuance System.
vcommerce.util.prompt.Playable
void appendTo(PromptPlayer player)
TTSPrompt
java.lang.String getText()
void setText (java.lang.String text)
Using TTS in a VoiceXML application
29
Note: The TTSPrompt class described in SpeechObjects on page 28 does not
work with the CorePromptPlayer class. You must use the SpeechChannel
implementation provided by the SpeechObjects framework to use TTSPrompt.
RCEngine
The RCEngine is a C++ recognition client-level API targeted for use by
developers who are integrating Nuance speech technology with an existing
telephony platform. To play TTS prompts through a Nuance RCEngine, you use
the same RCEngine PlayPrompts function you use to play a prerecorded
prompt. Instead of specifying a set of audio file names, you can specify a TTS
string by using the prefix "-tts_text:". This indicates that the following text is
the transcription for a TTS prompt, and not the name of a prerecorded audio file.
For example:
char const * prompts[2] = {"-tts_text:Hello world", NULL};
unsigned play_id = rce->GetUniqueID();
rce->PlayPrompts(prompts, play_id);
Note that the argument passed to PlayPrompts is actually a null-terminated
array of strings representing prompts to be played. This means you can include
multiple TTS prompt specifications, or mix TTS prompts with prerecorded
prompts. You can also add a pause to the array of prompts by using the prefix
"-pause:" followed by a number of milliseconds. For example:
char const * prompts[4] = {"welcome.wav", "-pause:500",
"-tts_text:Mister Jones", NULL};
unsigned play_id = rce->GetUniqueID();
rce->PlayPrompts(prompts, play_id);
See the Nuance Platform Integrators Guide shipped with the Nuance System for
more information on the RCEngine API.
You can use Vocalizer with the Nuance Voice Web Server to generate TTS
prompts from VoiceXML applications.
Vocalizer supports the following markup languages:
VoiceXML 1.0, in accordance with the W3Cs May 5, 2000 Note

(www.w3c.org/TR/voicexml)
SSML 1.0, based on the W3Cs April 5, 2002 Working Draft for Speech
Synthesis Markup Language Specification
(www.w3.org/TR/2002/WD-speech-synthesis-20020405)
Developers Guide
30

Text input with markup elements must be well-formed in order to be recognized
as XML markup rather than plain text. For example, the following text is valid
SSML:
<speak> Hello <break/> World</speak>
However, this text:
Hello <break/> World
would be synthesized as:
Hello angle bracket break slash angle bracket world.
Note that attributes specified as required in VoiceXML 1.0 and SSML 1.0 are not
enforced by Vocalizer. If a required attribute is missing from a tag, Vocalizer still
attempts to parse and handle that tag in a best-effort manner. Both VoiceXML 1.0
and SSML 1.0 translators treat all numeric attribute values as integers.
VoiceXML 1.0 elements
When using VoiceXML 1.0 elements, make sure Vocalizer is running in vxml
mode:
> vocalizer lm.Addresses=hostname:8471 -text_type vxml
The primary VoiceXML element for specifying a TTS prompt is the <prompt>
element. Any character data within a <prompt></prompt> block that is not
nested in another child element is treated as text to be synthesized by the TTS
server. For example:
<prompt>
Welcome to the Acme Travel Company.
</prompt>
Note: Text sent to Vocalizer must be well-formed XML in order to be recognized
as XML markup rather than plain text.
VoiceXML defines a number of child elements you can nest within a <prompt>
element to control the characteristics of the synthesized text, including those
listed in Table 1.
Table 1: Nested elements
Element Description
<break> Inserts a pause.
<emp> Increases or reduces the emphasis put on specific words.
<div> Allows you to specify either a sentence or a paragraph.
31
<audio> Plays an audio file within a prompt. Supports the following encoding:
Supports the following encoding:
For WAV files:
8 kHz, mulaw, unheadered
8 kHz, mulaw, Sphere header
8 kHz, PCM, Windows WAV format

For AU files:
8 kHz, mulaw (type 1), AU header
8 kHz, PCM (type 3), AU header

See URI formats for the SSML <audio> element on page 40.
<pros> Changes prosodic characteristics of synthesized speech, including
volume, pitch, range, and rate (speed).
<sayas> Provides a number of attributes for controlling how text is
synthesized.
For example, you can define a phrase to spell out an acronym or
pronounce an unusual word:
<prompt>
Contact the
<sayas sub="world wide web
consortium">W3C</sayas>
for more information.
</prompt>
This generates the prompt:
Contact the World Wide Web Consortium for more information.
<value> Inserts the current value of a variable into the string to be sent to the
TTS engine.
This shows a very simple example:
<var name="x" expr="12"/>
<prompt>
You have <value expr="x"> new email messages.
</prompt>
In a real application, of course, you would determine the value of the
variable programmatically rather than defining a static value.
Table 1: Nested elements (continued)
Element Description
Developers Guide
32

For more information on writing VoiceXML applications, see the Nuance Voice
Web Server documentation.
Note: The elements listed above may not all be supported on all versions of the
Voice Web Server.
SSML 1.0 elements
When using SSML 1.0 elements, make sure text input to Vocalizer is well-formed
XML.
Note: Unless you are using the <speak>, <paragraph>, or <sentence> tags, do
not include punctuation marks that denote the end of a sentence (., ?, !, or
...) in tagged data, as these cause the SSML element to be ignored.
Table 2: Supported SSML 1.0 elements
Element Attribute Value Notes
<speak> Root element.
Note: Only a single <speak> element is allowed per
request as this is a root element.
<paragraph>
<p>
Identifies the enclosed text as a paragraph.
<sentence>
<s>
Identifies the enclosed text as a sentence.
<phoneme> alphabet native
ipa
worldbet
xsampa
Provides a phonetic pronunciation for the enclosed
text. To use Vocalizers native phoneme set, change
the value of alphabet to native. See Resolving
XML parsing errors with predefined entities on
page 39.
Default is ipa
ph Specifies the phoneme string.
<sub> alias Replaces contained text with the value of the alias
attribute.
The alias attribute is required. If it is missing, an
error is logged and the <sub> element is ignored.
Note: This element replaces the sub attribute of the
<say-as> element, which is no longer supported.
33
<emphasis> level strong
moderate
reduced
none
Specifies that the contained text be spoken with
emphasis.
Default is moderate
<break> size small (500ms)
medium (1000ms)
large (2000ms)
none
Controls the pausing or other prosodic boundaries
between words. For example:
<speak>
Take a deep breath <break/> then
continue.
</speak>
Default is medium
Note: The <break> element must be enclosed by
surrounding text when used as a child of the
<audio> element.
time Indicates the duration of the pause in milliseconds
or seconds, for example: 2000ms, 2000 or 2s.
<say-as> Specifies the type of text construct contained within
the element.
type acronym Indicates contained text is an acronym.
Vocalizer capitalizes the contained text and
attempts to look it up in the acronym dictionary. If no
suitable match is found, the acronym is spoken
letter by letter.
type spell-out Indicates contained text should be spoken letter by
letter.
type number
number:ordinal
Ignores all non-digit characters except Roman
numerals.
Interpretation of intermixed Roman numerals and
digits is based on the first character. If the first
character is a Roman numeral, the entire sequence
is treated as a Roman numeral. Otherwise,
Vocalizer ignores the non-digit characters and treats
the data as a digit sequence.
type number:digits Ignores all non-digit characters.
type date Indicates contained text is a date.
Table 2: Supported SSML 1.0 elements (continued)
Developers Guide
34

<say-as> date:ymd
date:mdy
date:dmy
date:ym
date:my
date:md
date:y
date:m
date:d
Specifies the expected order for the components of
the date input.
For example:
<say-as type="date:ymd">
02/03/04</say-as>
and
<say-as type="date:dmy">
04/03/02</say-as>
would both be interpreted as:
March fourth two thousand and two
if using North American English voice packs.
type time:hms
time:hm
time:h
Supports data in a 24-hour clock format only. Data is
converted into 12-hour clock time. For example:
13:30:10 is converted to
one thirty and ten seconds p.m.
twelve oclock noon
twelve oclock midnight
type currency Supports values which can be interpreted as dollars
and cents only. International currency indicators,
such as USD or CAD, are not supported.
type name Indicates contained text is a proper name.
type net:email Indicates contained text is an internet identifier.
net:uri
type telephone Supports North American style phone numbers of 7,
10, and 11 digits. All punctuation is stripped and
ignored.
35
<say-as> type voicexml:date Correspond to the VoiceXML 2.0 built-in grammar
types.
For details on the required input formats for these
types, see the VoiceXML 2.0 specification at
www.w3.org/TR/2001/WD-voicexml20-20011023/
#dml2.3.1.1
voicexml:time
voicexml:boolean
voicexml:phone
voicexml:currency
voicexml:digits
voicexml:number
<prosody> Specifies prosodic information for the enclosed text.
For all <prosody> attribute tags, if user-specified
values are above or below the supported range of
values, the maximum or minimum value is
substituted for the user-specified value.
pitch Supported values:
Descriptive:
high (190)
medium (155)
low (140)
Absolute hertz: 50 to 300
Relative hertz: -50 to +100
Relative percentages: -30% to +30%
Relative semitones: +5st to -5st

Note: If % or st is omitted, the value is interpreted as
relative hertz. If - or + is omitted, the value is
interpreted as absolute hertz, even if % or st is
present.
Default is 155
Developers Guide
36

<prosody> contour Specifies the pitch contour for the contained text
using the syntax:
<prosody contour="(position,target)
[(position,target) ... ]" >
Where:
position is the percentage of the period of the

contained text, where "0%" corresponds to the
beginning of the contained text and "100%"
corresponds to the end.
target is the target pitch at that position. See

the description of the pitch attribute for
supported target values.
For an example of using contour, see the
Prosody and contour on page 38.
range Supported values:
Descriptive:
high (100)
medium (50)
low (10)
Absolute hertz: 0 to 100

Note: If relative values are specified, the attribute is
ignored.
Default is 50
rate Supported values:
Descriptive:
fast (80)
medium (100)
slow (130)
Fixed percentages: 75% to 150%

Note: If the % character is omitted, the value is
ignored.
Default is 100
37
SSML sample text
The following text demonstrates how SSML tags can be used to modify TTS
output. If you are running Vocalizer in ssml mode, you can hear the results by
copying the following text into the Offline Audio Generator (see Chapter 7).
Note: The text below can be copied in its entirety. If you want to test individual
paragraph samples, make sure to enclose the text in a <speak></speak> block
so that it is well-formed XML.
Say-as <speak>This is a demonstration of the Vocalizer <say-as
type="acronym">TTS</say-as> engine, showcasing a range of SSML
compatible features.
<prosody> duration Specifies milliseconds or seconds, for example:
3000ms, 3000 or 3s.
volume Supported values:
Descriptive: loud, medium, soft, silent
Fixed percentages: 0% to 100%

Note: Note: If the % character is omitted, the value is
ignored.
<audio> src URI Lets you specify the audio files to be played and text
to be spoken, using the file: or http: formats.
Supports the following encoding:
For WAV files:
8 kHz, mulaw, unheadered
8 kHz, mulaw, Sphere header
8 kHz, PCM, Windows WAV format
For AU files:
8 kHz, mulaw (type 1), AU header
8 kHz, PCM (type 3), AU header

See URI formats for the SSML <audio> element
on page 40.
Developers Guide
38

Prosody and volume <prosody volume="60%"> We can change </prosody> <prosody
volume="100%"> the volume </prosody> <prosody volume="70%"> of
ranges of words </prosody>.
Emphasis We can also apply emphasis, for example: Do you like this one, or
do you prefer <emphasis level="strong"> this </emphasis> one.
Alternatively, we can add emphasis using a combination of volume
pitch and duration markup flags. <prosody volume="50%"> For
example, how do you like </prosody><prosody rate="150%"
contour="(0,+0)(50,+50)(90,-10)" volume="100%"> this
</prosody><prosody volume="50%" contour="(0,+0)(100,-10)"> one
</prosody>.
Prosody, rate, and
duration
We can change the duration of an utterance by using the <prosody
rate="150%"><emphasis level="strong"> rate </emphasis></prosody>
and <prosody rate="150"> duration </prosody> flags. For example,
<prosody rate="150%"> This sentence is spoken 50% slower than
usual </prosody>. <prosody duration="4s"> Whereas, this sentence
is exactly 4 seconds in duration </prosody>.
Prosody, pitch, and
phoneme
We can make absolute and relative pitch changes to the speech, for
example, <prosody pitch="170"> this sentence is spoken with an
average frequency of 170 hertz </prosody>. <prosody pitch="-20%">
This sentence's average frequency has been <phoneme ph="l o *r
d"/> by 20 Hertz </prosody>.
Prosody and contour An arbitrary pitch contour can be applied using the <emphasis
level="high"> contour </emphasis> flag. <prosody
contour="(0,-30)(90,+30)"> For example, this sentence
interpolates a relative frequency change from 30 hertz below
normal at the beginning to 30 hertz above normal at the
end</prosody>.
Misapplied elements Multiple target frequencies can be specified across the range of
the contour flag, allowing an unskilled user to easily create
prosodic monstrosities. For example, <prosody
contour="(0,+20)(70, +0)"> question </prosody> type intonations
can be badly <prosody rate="130%"
contour="(0,+40)(50,-10)(100,+90)"> applied </prosody>. However,
in skilled hands and applied responsibly, these flags provide a
user with an impressive level of prosodic control.</speak>
Resolving XML parsing errors with predefined entities
39
Resolving XML parsing errors with predefined entities
When using Vocalizer in vxml or ssml mode, certain characters may cause XML
parsing to fail so that no audio is rendered. For example, the Worldbet and
XSAMPA phonetic alphabets include characters which may cause parsing errors
if embedded within the ph attribute of the <phoneme> element.
To resolve parsing errors, use XML predefined entities to represent these special
characters, as shown in the table:
As an example, the XSAMPA alphabet uses g_< to indicate an implosive g.
However, the < and > characters are reserved in XML-based markup
languages, so the following text will be parsed incorrectly:
I like <phoneme ph="g_<oldfIS" alphabet="xsampa">goldfish
</phoneme>.
To resolve the parsing error, you would replace < with < as follows:
I like <phoneme ph="g_<oldfIS" alphabet="xsampa">goldfish
</phoneme>.
See the W3Cs website for more information on predefined entities
(www.w3.org/TR/2000/REC-xml-20001006#dt-escape).
Encoding
By default, Vocalizer markup processors expect any XML markup to be UTF-8
encoded, per RFC 2279. A request without an XML header must be encoded in
UTF-8. For example:
<speak> This message should be in UTF-8 </speak>
For non-English languages with accented characters, text requires UTF-8
encoding of the ISO-8859-1 characters before being sent to Vocalizer. In SSML
Reserved character Replace with predefined entity:
& (ampersand) &
(double quote) "
< (less than) <
> (greater than) >
(apostrophe) '
Developers Guide
40

mode, Vocalizer can also interpret ISO-8859-1 input character if you provide an
appropriate XML header:
<?xml version="1.0" encoding="ISO-8859-1"?>
<speak> This message is in iso 8859 1 format </speak>
For most English text, encoding is not an issue because of the direct
correspondence between UTF-8 and 7-bit ASCII; characters numbered 0 to 127
in 7- or 8-bit ASCII are the same as the characters numbered 0 to 127 in UTF-8.
However, if you enter non-English expressions in SSML mode, such as words
like coup d'tat, you must specify the encoding. Failure to enter the encoding
causes unpredictable behavior. For example, entering text in the Offline Audio
Generator with at least one accented character produces no output.
Caution: When using non-English voice packs and entering text in SSML mode,
you must specify the encoding type, ISO-8859-1, to avoid unpredictable
behavior. For example, entering text in the Offline Audio Generator with at least
one accented character produces no output. See Text encoding on page 88.
The following encoding values are recognized by Vocalizer:
utf-8
utf-16
iso-8859-1
us-ascii
When using the <phoneme> element in SSML, if you select IPA as the alphabet,
the IPA characters can be represented by multi-byte values only, and the entire
document passed to Vocalizer must therefore be encoded using either UTF-8 or
UTF-16.
URI formats for the SSML <audio> element
Vocalizer handles a specific set of http: and file: URI formats (per RFC s 2616 and
2396) for the VoiceXML <audio> element. For example, this table shows illegal
formats:
URI Error
C:/foo.wav Invalid format per RFC 2396
file://C:/foo.wav Uses only two forward slashes (/) instead of three
between the colon (:) and the C.
The correct form is: file:///C:/foo.wav
URI formats for the SSML <audio> element
41
Table 3 and Table 4 demonstrate the correct formats for absolute and relative
URIs, and show how these URI formats are handled by Vocalizer.
Some notes:
You can use either / or \ for path components
The only authority Vocalizer handles for file URIs is localhost or

127.0.0.1
Syntax marked as Unix will not work on Windows; examples containing

C: will not work on Unix
The base URI for relative file URIs is:

file://localhost/%VOCALIZER%/data/audioFiles/
This means:
If the scheme is not specified it will be assumed to be file:
If the authority is not specified, it will be assumed to be localhost
If the path portion of the URI is relative, the file is searched for in the
%VOCALIZER%/data/audioFiles/ directory
Table 3: Absolute URIs
URI Reference
file://localhost/C:/foo/bar.wav C:\foo\bar.wav
file://localhost/foo/bar.wav /foo/bar.wav (Unix)
file:///C:/foo/bar.wav C:\foo\bar.wav
file:///foo/bar.wav /foo/bar.wav (Unix)
file:/C:/foo/bar.wav C:\foo\bar.wav
file:/foo/bar.wav /foo/bar.wav (Unix)
http://localhost/bar.wav
http://audio-server/audiofiles/bar.wav
http://audio-server:8080/audiofiles/bar.wav
Table 4: Relative URIs (file: format only)
URI Reference
//localhost/C:/foo/bar.wav C:\foo\bar.wav
Developers Guide
42

//localhost/foo/bar.wav /foo/bar.wav (Unix)
///C:/foo/bar.wav C:\foo\bar.wav
///foo/bar.wav /foo/bar.wav (Unix)
C:/foo/bar.wav / C:\foo\bar.wav
/foo/bar.wav /foo/bar.wav (Unix)
foo/bar.wav %VOCALIZER%\data\audioFiles\foo\bar.wav
./foo/bar.wav %VOCALIZER%\data\audioFiles\foo\bar.wav
../foo/bar.wav %VOCALIZER%\foo\bar.wav
bar.wav %VOCALIZER%\data\audioFiles\bar.wav
./bar.wav %VOCALIZER%\data\audioFiles\bar.wav
../bar.wav %VOCALIZER%\bar.wav
Table 4: Relative URIs (file: format only)
URI Reference
Chapter 5 Starting Vocalizer through the Launcher
Using the Vocalizer Launcher
5
43
Chapter 5
Starting Vocalizer through
the Launcher
The Vocalizer Launcher is graphical tool that allows you to create, edit, and save
start-up configurations for running Nuance Vocalizer.
From the Vocalizer Launcher interface, you can configure and start vocalizer,
along with the nlm process. You can also start any of the other GUI tools
provided with Vocalizer.
To use the Launcher:
1 From the Windows Start menu, go to Programs Nuance Nuance Vocalizer
3.0 Executables Vocalizer Launcher
The Configurations window lists the voice packs and configurations
available on your system.
Developers Guide
44

2 After selecting a configuration, choose either:
StartOpens the Status window and starts vocalizer with the default
options for the selected configuration.
NewOpens the Configuration Editor for you to create a new

configuration.
EditOpens the Configuration Editor and allows you to modify the

selected configuration.
DeleteRemoves the selected configuration.
CopyDuplicates an existing configuration, which you can then open

and modify. This is useful if you want to create a similar configuration
with only a few adjustments.
Configuration editor
The configuration editing window lets you specify your configuration options
that you would normally set on the command line.
The window groups configurations options into these sections:
Chapter 5 Starting Vocalizer through the Launcher
45
Vocalizer configuration
Configuration nameAllows you to rename your configuration
Channel countAllows you to enter the number of channels you want to

use, up to the maximum specified in your license
Vocalizer portAllows you to change the TTS port setting for audio
generation
Dictionary portAllows you to change the dictionary port setting for

phonetic generation.
License manager
License managerAllows you to enter and select the license manager

running your nlm process
New/EditOpens a dialogue window in which you can either specify a

different license manager host and port to run your nlm process or start
the local license manager
DeleteRemoves the license manager running your nlm process
Resource manager
Connect to a resource managerEnables connection to a resource

manager
Resource managerSpecifies hostname and port setting
Use resource nameSets the tts.ResourceName parameter to the name

by which your Vocalizer server identifies itself
Voice configuration
PersonaLists the voice packs available from which you can make a
selection
Audio encodingLets you choose from mulaw or alaw audio encoding,

depending on the voice pack installed
Voice sizeDisplays the type of amount of compression, depending on

the voice pack installed
Text and filter configuration
Text typeLets you select Plaintext, VXML, or SSML
FilterAllows you to select the text filter (or filters) you want to use and
specify the order in which they are applied
Developers Guide
46

Additional command-line optionsAccepts a single command line option

(with no white spaces). For example, you can enter the -preallocated_licenses
or -voice_from_disk option.
Note: Do not enter options that take arguments (such as -text_type ssml)
because white spaces are not accepted in this field.
You can name the set of configuration options and save them for reuse.
Status window
Selecting Start from the main window opens the Status window, which displays
the current state of the vocalizer process.
Once the vocalizer process has been started, you can start these GUI tools:
Offline Audio GeneratorSee Chapter 7
Email Configuration ToolSee Chapter 6
Filter EditorSee Chapter 6
Dictionary EditorSee Chapter 8

Closing this window stops the vocalizer process.
Chapter 6 Preprocessing text input
Passing email messages to Vocalizer
6
47
Chapter 6
Preprocessing text input
Vocalizer allows you to control the way email messages and other text input is
processed. This chapter explains how to do this using the Email Reader
Configuration tool and the Text Replacement Filter Editor.
Vocalizer can parse and read email messages encoded using the RFC 822
standard. To enable this feature, add the -filter option on the vocalizer command
line:
-filter email
This allows you to pass email message text directly to Vocalizer.
Note: Vocalizer currently cannot parse email in Encrypted, MIME, or any type of
8- to 7-bit character encoding such as quoted-printable, UUEncode, or base 64
encoding.
Features of the Email Reader Configuration Tool
You can use the Email Reader Configuration Tool to configure the way email
messages are converted to speech and save your configuration options. The
email handler is able to switch between multiple configuration files.
The Email Reader Configuration Tool allows you to do the following:
Load, edit, and save email configurations
Specify which email fields to read and in what order
Specify text to be spoken before and after each field

Developers Guide
48

Specify a maximum number of names to be read from the To and Copied

fields
Normalize email addresses and remove separator lines from signatures

The Email Reader is a Java application built on Java 2. The Java Runtime
Environment (JRE) 1.3 is installed with Vocalizer to allow you to run this utility.
See Third-party software requirements on page 4.
Using the Email Reader Configuration Tool
To start this tool:
On Windows, from the Programs group in your Windows Start menu, select
Nuance Nuance Vocalizer 3.0 Executables Email Reader
Configuration Tool
On Solaris, enter:
> EmailReaderConfigTool.sh
49
The Text In box displays a sample RFC 822 email. The Text Out box allows you
to see the results of configuration choices, displaying how the email messages
will be read. Use the tabs to choose which field you want to edit. You can see
your updates in the Text Out box as you make your changes.
Specifying the fields
to read
Under each tab is a checkbox labeled Read field_name field. Check the Enabled
box to specify whether or not you want information from that field included in
the synthesized output. For example, under the Date tab in the previous
graphic, Read Date field is enabled.
Specifying order The Field Read Order list box lists all the email fields in the order they will be
read. To change the order, select the field you want to move, then click the Up or
Down button.
Specifying text
output
Each field of the email can be prefaced or followed by user-defined comments.
To edit this text, select the tab for the particular field, either Date, From, To,
Copied, Subject, or Body. Enter your comments in the appropriate text box.
The To and Copied fields often include several names. You can decide if you
want Vocalizer to tell you the total number of names listed, and if you want
Vocalizer to read each name.
Developers Guide
50

To avoid having a long list of names synthesized, you can specify a maximum
number, X, so that Vocalizer only reads the list if there are fewer than X names.
For example, if you set the If there are less than field to 5, the application reads
all the names that are on the list only if there are fewer than five. If there are five
or more names, nothing is read.
If you want Vocalizer to tell you the total number of names listed, specify a
minimum number, Y, so that Vocalizer only tells you the total when there are
more than Y names. For example, if you set the If there are more than field to
1, the application tells you the total number of names only if there is more than
one name. If there is only one or zero names, nothing is said.
Address and
signature handling
From the Miscellaneous tab, you can enable the settings for normalizing email
addresses and signature handling.
Normalizing addresses uses white space to replace underscore characters (_)
and to separate digits and characters within a strings. For example:
Address input Normalized result
[email protected] a [email protected]
[email protected] ted [email protected]
51
Some email signatures include a separator line. Check the Remove non
alpha/numeric characters? to remove lines which do not contain alpha or
numeric characters.
Distributing changes
You can edit the email configuration files for all your Vocalizer servers on a
single development machine. When you save changes in the Email Reader
Configuration Tool, the configuration file is compiled and put in the
%VOCALIZER%\data\locale\email directory. You can distribute this file to
your other TTS server machines by copying the contents of the
%VOCALIZER%\data\locale\email directory to the same path on the
machine(s) where you want to distribute the changes.
Developers Guide
52

Creating custom filters for text replacement
The Text Replacement Filter Editor allows you to create custom filters to modify
how specific text input is rendered by the TTS engine.
Text replacement filters allow you to specify a set of related translations,
mapping text input to a replacement value. For example, the emoticon filter
includes several replacements, such as:
Each filter is stored in a single file. There is no limit to the number of filters you
can create and use.
To enable this feature, add the -filter option on the vocalizer command line:
lm.Addresses=localhost:8471 -filter your_custom_filter
Applying filters
You can specify multiple filters and choose the order in which they are applied.
Enter the filter names using a comma-separated list with no spaces. For
example, if you have a filter named medical, which expands medical acronyms,
you can include it with the emoticon filter like this:
lm.Addresses=localhost:8471 -filter emoticon,medical
Filters are applied in the order in which they appear. The order specified can
affect the output if replacement text generated by one filter matches a transform
for a following filter. For example, if you input this text:
My computer is broken. :-(
the emoticon filter outputs:
My computer is broken. sad face
If your medical filter recognizes sad as an acronym to be expanded, then it
would output:
My computer is broken. Seasonal Affective Disorder face
Reversing the order of the filters yields the following output:
Input Replacement text
:-) smiling face
;-) winking face
:-( sad face
Creating custom filters for text replacement
53
My computer is broken. sad face
Using the Text Replacement Filter Editor
To use the Text Replacement Filter Editor:
1 Start the tool:
On Windows, from the Programs group in your Windows Start menu,

select Nuance Nuance Vocalizer 3.0 Executables Text Replacement
Filter Editor
On Solaris, enter:
> ReplacementFilterEditor.sh
2 In the Replace text box, enter the text you want recognized. You can also
choose to:
Match whole word only

By default, Vocalizer tries to match the text string as a word, and looks for
word delimiters, including spaces, sentence terminators, hyphens, and
parentheses. If the text string is found within another word, it is ignored.
If you want the text string to always be replaced, uncheck this box. For
example, if you create a filter to block obscenities, you would probably
want the text filtered, even if it appears as part of other text strings.
Match case
By default, Vocalizer ignores the case for text entered. Selecting Match
Case allows a case-sensitive search for the specified string.
3 In the With text box, enter the text you want Vocalizer to output.
4 Click Add Replacement.
Developers Guide
54

When youre done, save the list of replacements as a file in the
%VOCALIZER%\data\replace-filter directory. You can use the Text Replacement
Filter Editor to open and edit any of your text replacement files.
Distributing changes
You can create the filter files for all your Vocalizer servers on a single
development machine. After saving your changes in the Text Replacement Filter
Editor, the filter files are stored in the %VOCALIZER%\data\replace-filter
directory. You can distribute this file to your other TTS server machines by
copying the contents of the %VOCALIZER%\data\replace-filter directory to the
same path on the machine(s) where you want to distribute the changes.
Chapter 7 Generating audio files from text input
Using the Offline Audio Generator
7
55
Chapter 7
Generating audio files from
text input
Vocalizer can generate audio streams directly from user input or from text files.
The resulting audio stream can be saved to files for playback. This chapter
explains how to do this using the Offline Audio Generator.
The Offline Audio Generator is a GUI tool which accepts text input either
directly from users or from text files.
The text input is sent to Vocalizer, which returns an audio stream. The audio is
played back to the user and can be written to a file in one of these formats:
Headerless mulaw
Mulaw with Sphere Header (playable with Nuances playwav utility)
Headerless Linear
Linear with Sphere Header (playable with Nuances playwav utility)
Linear Riff (Windows default .wav format)

Developers Guide
56

.
To use the Offline Audio Generator:
1 Make sure the vocalizer process is running.
2 Start the tool, either using the Launcher (see Chapter 5) or:

select Nuance Nuance Vocalizer 3.0 Executables Offline Audio
Generator
On Solaris, enter:
> OfflineAudioGenerator.sh
The GUI will appear.
3 Enter your IP address and the TTS port number. (The default TTS port is
32323).
4 Specify the text to be rendered by selecting either:
Chapter 7 Generating audio files from text input
57
Use this textEnter text in the input field
Use text from fileSelect a text file through the Browse button
5 Choose whether or not to save audio output to a file. If you select Save
output to file:
a Specify the file to which audio output should be written, by either
entering the filename in the output file field or selecting an existing audio
file to overwrite.
b Specify the audio format for saving output from the drop-down menu.
By default, audio files are generated as .wav files. Choose the Sphere
headered for use with Nuance applications, or headerless formats for use
with other audio editors.
6 Click Play.
You can stop playback at anytime by clicking the Stop button. The Exit
button also stops playback before quitting.
Validating multi-server configurations
If you want to test the Vocalizer installation on other servers in your
configuration, enter the IP address of the appropriate machine. The text is sent
to the Vocalizer server on that machine.
Developers Guide
58

Chapter 8 Customizing your applications dictionary
Features of the Dictionary Editor
8
59
Chapter 8
Customizing your
applications dictionary
Vocalizer includes the Dictionary Editor, a graphical tool that allows you to
dynamically modify the dictionary files used by the TTS engine.
This chapter describes how to use the Dictionary Editor to add words and
acronyms to your dictionary, or modify the pronunciation of existing entries. It
also explains how to distribute changes made on one machine to other Vocalizer
servers.
The Dictionary Editor is a Java application built on Java 2. The Java Runtime
Environment (JRE) 1.3 installed with Vocalizer allows you to run this utility. See
Third-party software requirements on page 4.
The Dictionary Editors main menu includes File and Options:
From the File menu you can either:
Save Changes LocallySaves your modifications to your machine only
Save Changes to ServerCompiles your changes and uploads them to

Vocalizer
Note: Selecting Save Changes to Server blocks all TTS requests during the
time it takes to compile the changes, so this option should not be used
during peak periods. You can only use this option if the edited dictionary
files are for the same language as the voice pack you are running.
From the Options menu, you can either:
Modify the port settings for Vocalizer audio generation. The default TTS
port is 32323, and the default port for phonetic generation is 22552.
Developers Guide
60

Select the dictionary file to modify according to language, either:
North American English
U.K. English
Australian/New Zealand English

By default, North American English language dictionary files are loaded.
Because dictionary files are language-specific, if you are running a North
American English voice, for example, but modify the U.K. English
dictionary, you can save your changes, but you will only hear the effect
the next time you run a U.K. English voice pack.
The main window of the Dictionary Editor has two tabbed panes:
Word EditorLets you view the current dictionary entries, add new words,
modify the way existing entries are pronounced, or remove entries
Acronym EditorLets you view all acronyms known to Vocalizer in the

dictionary entries, enter new acronyms, modify the way existing acronyms
are expanded, or remove entries
61
Depending on which tabbed pane you have selected, clicking New or Edit in the
Word Options section opens either the Word Editor or Acronym Editor dialogue
window. Selecting Delete removes whichever entry is currently selected in the
Dictionary Entries.
Note: If you close the Dictionary Editor without saving, your changes will not
take effect.
Developers Guide
62

Audio feedback
Vocalizer provides immediate feedback on your entries in the Word Details and
Scratchpad sections.
The System Pronunciation field displays the current phonetic translation for a
selected word, and you can use the Play button next to that field to hear the TTS
engines pronunciation. When you change the pronunciation of an existing
entry, you can hear your modified version, using the Play button next to the
User Pronunciation field.
The Scratchpad lets you enter a text string using a dictionary entry, then play
either the user-specified pronunciation or the systems pronunciation in that
context. See Audio scratchpad on page 64.
63
Word details
The Word Details section defines the usage and phonetic characteristics of your
entry through the following fields:
SpellingDisplays the entry
Part of speechDisplays the entrys linguistic role
Say As (in the Acronym Editor only)Displays the text expansion for
acronyms
User PronunciationDisplays the user-defined phonetic translation
System PronunciationDisplays the systems phonetic translation

Phoneme set
Pronunciation is built with phonemes. Vocalizers phonetic translation is based
on the Computer Phonetic Alphabet (CPA). The CPA is also used by the Nuance
System to build pronunciations in Nuance recognition dictionaries.
The Phoneme Set contains a complete list of phonemes, along with stress
markers and a syllable separator, which you can use to build pronunciations for
your entries. You can use the Play button beside each phoneme to hear the
systems pronunciation and determine your selection. Clicking on the phoneme
inserts it at the current cursor position of the User Pronunciation.
Auto-phonetic
transcription
When you enter new words and new expansions for acronyms in the Say As
field, Vocalizer automatically provides a phonetic translation for your entry.
Based on the TTS engines pronunciation, you can modify the phonetic
translation appropriately.
Inserting stress
markers
In addition to the phonetic translation, your pronunciation definition can
include stress markers for improved prosody:
Enter stress markers immediately preceding the syllable to be accented. For
example:
Marker Symbol
Primary Stress Single quote ()
Secondary Stress Period (.)
Syllable separator Dash (-)
Developers Guide
64

Audio scratchpad
The Scratchpad lets you send text to Vocalizer to test pronunciations in the
context of your application. You can enter text strings in which your modified
entries may be used, then click the Play button below the Scratchpad to hear
your text synthesized.
Choosing either User or System determines if the Scratchpad plays your entry
according to the systems pronunciation or the phonetic translation youve
entered.
See Testing your entries in Scratchpad on page 67.
Using the Dictionary Editor
To use the Dictionary Editor:
1 Make sure the vocalizer process is started.
2 Start the Dictionary Editor:

select Nuance Nuance Vocalizer 3.0 Executables Dictionary Editor
On Solaris, enter:
> DictionaryEditor.sh
Note: Changes are not saved automatically when you close the editing tool. After
following the procedures for adding or modifying dictionary entries, make sure
to save your changes before closing the Dictionary Editor.
Word Pronunciation encoding
emoticon i ' m o - t I . k A n
laden ' l A - d * n
65
Enabling logging
You can edit the Dictionary.bat file to enable debug information to be sent to a log
file (dictionaryEditor.log) in the %VOCALIZER%\bin\win32 directory:
1 From the %VOCALIZER%\bin\win32 directory, open Dictionary.bat as a text
file.
2 Before -D"INSTALL_PATH", add the parameter setting -D"debug=true". For
example:
"C:\Program Files\JavaSoft\JRE\1.3\bin\java" -jar -cp
"C:\Program Files\Nuance\Vocalizer3.0\java"
-D"debug=true"
-D"INSTALL_PATH=C:\Program Files\Nuance\Vocalizer3.0"
"C:\Program Files\Nuance\Vocalizer3.0\java\
DictionaryEditor.jar"
Adding new entries
To add a new word or acronym to the dictionary:
1 In Word Options, click New to open the Word Editor or Acronym Editor.
2 In the Spelling field, enter the new word or acronym.
Make sure to enter the correct spelling for text transcriptions sent to the TTS
engine; dictionary entries are not case-sensitive.
If you are entering a new word, the User Pronunciation field will
automatically display a phonetic translation based on the text in the Spelling
field.
3 From the Part of Speech drop-down menu, make the appropriate selection,
for example, noun, proper or adverb.
4 If you are entering a new acronym, in the Say As field, enter the expansion.
For example, with the acronym TTS you would enter text-to-speech.
The User Pronunciation field automatically displays a phonetic translation
based on the text in the Say As field.
5 In the User Pronunciation, build or modify the phonetic pronunciation for
your word:
a This field is filled automatically, but you can modify the phonetic
translation using the Phoneme Set. See Phoneme set on page 63.
b Add Primary or Secondary Stress symbols before the syllable to be
accented; see Inserting stress markers on page 63.
Developers Guide
66

6 Press the Play button next to User Pronunciation or use the Scratchpad to
test your phonetic translation; see Testing your entries in Scratchpad on
page 67.
7 When you are happy with the pronunciation, click OK to add the word to the
dictionary.
8 Save your changes. From the File menu, select either:
Save Changes LocallySaves your changes, but does not dynamically

update Vocalizer.
Save Changes to ServerSaves and dynamically updates Vocalizer.

You can only use this option if the edited dictionary files are for the same
language as the voice pack you are running.
Modifying dictionary entries
To modify the pronunciation associated with an existing dictionary entry:
1 Select one of the Dictionary Entries:
From the Word Editor, you can modify word pronunciation
From the Acronym Editor, you can modify acronym expansion and
pronunciation
2 In Word Options, click Edit to open up the Word Editor (or Acronym Editor).
3 Modify the appropriate Word Details, as described in Adding new entries
on page 65.
4 Click OK when you are done.
SaveSaves your changes, but does not dynamically update Vocalizer.
Save and UploadSaves and dynamically updates Vocalizer.

67
Testing your entries in Scratchpad
When you add or edit a dictionary entry, you can use the Scratchpad to test your
modifications:
1 Select a dictionary entry to edit; phonetic translation for the entry appears in
both the User Pronunciation and System Pronunciation fields.
2 Edit the User Pronunciation.
3 In the Scratchpad, enter a text string including the word you have edited.
4 In the Pronunciation Source, select User or System.
Select User to play your text string using the modified pronunciation; if
youve changed the Say As field for an acronym, Scratchpad plays the
modified expansion.
Select System to play your text string using the systems version of that
entrys pronunciation.
Deleting entries
To delete from the Dictionary Entries:
1 In Dictionary Entries, select the appropriate word or acronym.
2 In Word Options, select Delete.
SaveSaves your changes, but does not dynamically update Vocalizer.
Save and UploadSaves and dynamically updates Vocalizer.

You may be able to increase synthesis time by removing unused words from
your dictionary.
If you remove all Dictionary Editor entries, all pronunciations will be
determined solely by the TTS engine.
Developers Guide
68

Distributing dictionary changes
You can edit the user dictionaries for all your Vocalizer servers on a single
development machine. When you save changes in the Dictionary Editor, the
dictionary files are compiled and put in the
%VOCALIZER%\data\locale\Lexicon directory. You can distribute these files
to your other server machines as follows:
1 From the machine on which your edits were made, copy the contents of the
%VOCALIZER%\data\locale\Lexicon directory to the same path on the
machine(s) where you want to distribute the changes.
2 Optional. The dictionary source files are in the
%VOCALIZER%\data\locale\dictionary directory. If you wish to keep the
source files on each of the machines to which changes were distributed, copy
the contents of %VOCALIZER%\data\locale\dictionary to the same path
on the destination machine(s). These files are not used during regular
Vocalizer server execution.
Chapter 9 Techniques for enhancing audio output
Phrasing and punctuation
9
69
Chapter 9
Techniques for enhancing
audio output
The layout, punctuation, and capitalization of text sent to Vocalizer affect
speech output. Certain types of characters and phrasesfor example, digits,
dates, and email addressesare handled in particular ways. By changing the
way you transcribe phrases and characters, you can alter the speech synthesis.
This chapter outlines how Vocalizer processes your text input.
This information applies to text processing for the English language in North
American, U.K., and Australian/New Zealand dialects. Note that date formats
for North America differ from U.K. and Australian/New Zealand formats. See
Dates on page 84.
For information on text processing for Canadian French, see Appendix A.
Vocalizer recognizes punctuation used to end phrases, sentences, and
paragraphs. Vocalizer strips all layout from input text and uses it, in conjunction
with punctuation, to determine phrasing, thereby enabling natural-sounding
intonation and pausing at the ends of phrases, sentences, and questions. Text
that is laid out and punctuated in a natural fashion, as it would be for normal
written text, will be read in a natural predictable way.
TIP: For the most natural-sounding speech, try to break very long sentences
into multiple sentences. Sending input such as turn left and go 1 mile turn
right and go 2 miles, will not sound right because there is no punctuation in
the input text.
Table 5 explains how Vocalizer interprets punctuation:
Developers Guide
70

Table 5: Interpretation of punctuation
Punctuation mark Interpretation
Comma (,)
Semicolon (;)
Colon (:)
End of phrase
Either a:
Period (.)
Question mark (?)
Exclamation mark (!)
Ellipsis (...)
Followed by a single space
End of sentence
Period (.)not immediately following an abbreviation or
a single characterfollowed by a single space
See Hyphens and dashes on page 71
End of sentence
Either a:
Period (.)
Question mark (?)
Exclamation mark (!)
Ellipsis (...)
Followed by a:
Single quote ( )
Double quote ( )
Closing parenthesis )
and either a single space, a new line, or the end of input
End of sentence
Period (.) preceded by a non-alphabetic character End of sentence
Two or more consecutive new lines End of paragraph
End of a sentence followed by a new line and either
another new line, a tab, or a space
End of paragraph
End of input End of paragraph
71
Note: The last character in a sentence does not necessarily have to be a sentence
terminator. In some cases, sentences are enclosed inside quoted or
parenthesized sections, for example:
Then he said: Sorry, I cant do that.
Periods
Periods can be used as sentence terminators, abbreviation markers, decimal
points, IP address dividers, and date dividers. Vocalizer determines the correct
interpretation from the context:
Sentence terminatorsSee Table 5 on page 70
Abbreviation markersThe period is considered an abbreviation marker if it

has not been identified as a sentence terminator and it follows either a single
character or a known abbreviation. See Abbreviations on page 73.
Note: As a general rule, allow two spaces to follow a sentence terminator, and
use a period and an optional space to terminate an abbreviation.
IP address dividersSee Periods in digit sequences on page 78
DecimalsSee Decimal fractions on page 79
Date dividersSee Dates on page 84

Hyphens and dashes
Since hyphens, dashes, and minus signs are represented by the same ASCII
character, ambiguous situations may occur at times. To avoid ambiguous
situations, Vocalizer uses a set of disambiguation rules. Specifically, Vocalizer
considers the - character to be:
A minus sign if it is immediately followed by a digit.

When a minus sign is identified, it is replaced by the word minus.
A dash if it has a whitespace on either side or if it is the first character in a

sentence.
The dash is removed, and the output is marked to indicate the punctuation.
A soft hyphen if it follows a letter and is followed by a carriage return and a

letter.
The words on either side of the hyphen are joined and handled as a single
word.
Developers Guide
72

A hard hyphen if it is used in any other situation, for example, if it appears

within a word.
Hyphenated words are split into constituent words. Also, the fact that
hyphenated words are part of a single unit is recorded by the system for later
reference and processing.
TIP: A hyphenated score, such as a football score of 37-27, should be sent to
Vocalizer as 37 to 27 or 37, 27so the output is not rendered as thirty seven
minus twenty seven.
Parentheses and double quotes
Vocalizer handles parentheses and double quotes. If a closing parenthesis or
double quote occurs without any preceding matching symbol, for example, an
open parenthesis, Vocalizer reads the close parenthesis or double quote as text.
If two double quotes appear consecutively, they will also be read as text. For
example:
Apostrophes and single quotes
Since apostrophes and single quotes are represented by the same ASCII
character, ambiguous situations may occur at times. For example, consider a
case where two or more words ending in s followed by an apostrophe ()
character are found consecutively in the same sentence:
Sorry, we do not deal with customers complaints.
In cases such as this one, Vocalizer considers the last () character as the closing
of a quoted section (if previously opened) and the remaining () characters as
apostrophes.
To avoid ambiguous situations, Vocalizer uses the following set of
disambiguation rules regarding the () character:
An apostrophe if the character appears within a word
An open single quote if the character is the first character of a word

Input text Output text
(hi) hi
hi) hi parenthesis
hi hi
quote quote
Abbreviations
73
A closing single quote if the character is the last character of a word whose last
letter is not s
An apostrophe if the character is the last character of a word whose last letter
is s and a quoted section has not been opened previously
Ignored in all other cases

Abbreviations
Abbreviations may consist of one or more alphabetic characters, which are
either all lowercase, all uppercase, or have an uppercase first letter followed by
lowercase letters. You can separate abbreviations from words by adding:
A period and a spaceFor example, Vocalizer, v. 3
A period onlyFor example, Vocalizer, v.3
A space onlyFor example, Vocalizer, v 3

Single-letter abbreviations, ambiguous abbreviations, homographic
abbreviations, and measurement abbreviations require additional work, as
described in the following subsections for detailed information.
Single-letter abbreviations
Ambiguous situations may occur between single-letter abbreviations and
initials. Where a possible single-letter abbreviation is found, Vocalizer interprets
the letter as:
An abbreviation if it is followed by one or more digits and it appears in the list

of abbreviations
An abbreviation if it follows one or more digits and it is a measurement or

numeric abbreviation
An initial in all other cases

Developers Guide
74

Ambiguous abbreviations
Ambiguous abbreviations and their long form must be specified in the
abbreviations file. At present, Vocalizer only disambiguates titles and road
types, for example St. (saint or street) and Dr. (doctor or drive). Vocalizer
resolves ambiguity by the following set of rules in the order prescribed:
1. The abbreviation is a title if it is the first word in a sentence.
2. The abbreviation is a road type if it is followed by a comma, a semi-colon, a
colon, or a period.
3. The abbreviation is a title if it follows a comma, a semi-colon, or a colon.
4. The abbreviation is a title if the previous word is a road type.
5. The abbreviation is a title if the first letter of following word is capitalized.
6. The abbreviation is a road type if the first letter of the previous word is
capitalized.
7. The abbreviation is a title in all other cases.
Homographic abbreviations
Homographic words are words that are spelled the same way but have different
meaning, for example, the noun lead and the verb lead. Homographic
abbreviations are abbreviations that are spelled the same way as a word or letter
but have different meaning, for example, the abbreviation Sat. and the verb
sat or in and in. (inch).
Vocalizer expands homographic abbreviations if they are followed by a period
and a single space, or a period and a punctuation mark, for example:
Homographic abbreviations can also be acronyms if they are written with at
least one uppercase letter (not including the first letter of the word). For
example, SUN and sUn can either be the abbreviation for Sunday or an
acronym.
sat. 4th dec saturday fourth december
sat.. Then he went saturday (pause) then he went
he sat. Then he went he sat (pause) then he went
Abbreviations
75
As mentioned earlier, Vocalizer expands homographic abbreviations if they are
followed by a period and a single space, or a period and a punctuation mark, for
example:
TIP: Be careful when creating a custom text replacement filters that replace
homographic abbreviations. For example, a filter that replaces me with
Maine could incorrectly render the word me when encountered in regular
text.
Measurement and numeric abbreviations
Measurement and numeric abbreviations are only recognized if they follow a
digit (with or without a space), for example, 1M and 4 cm.
With measurement abbreviations, the following rules apply:
If the preceding word is the number 1 (entered as a digit), Vocalizer uses

the singular version of the long form; otherwise, it uses the plural version.
When a number is followed by an alphabetic character, Vocalizer checks

whether the characters are specified in the list of measurement
abbreviations:
If the characters are specified, it expands the abbreviation to its long

form.
If the characters are not specified, Vocalizer treats the characters as

separate letters.
For example:
Input text Notes
SUN computers Interpreted as an acronym
SUN. 1st June Converted to Sunday because SUN is followed by a
period and a single space
we are coming on SUN..
Computers are here.
Converted to Sunday because SUN is followed by
two periods
sun.1st June Converted to Sunday because sun is all lowercase
3 kg three kilograms
3 cm three centimeters
Developers Guide
76

Acronyms and initials
Vocalizer also handles acronyms and initials. Vocalizer defines acronyms and
initials as follows so that the pronunciation tool can treat them appropriately.
A string of letters including one or more uppercase letters (not including the
first letter)
A string of letters separated by periods
A single letter, except where the letter is an abbreviation or I, i, A, or

a
Since I, A, and a can be considered as words or letters, Vocalizer
distinguishes them as follows:
I, A, or a is a letter if followed by a period and a single space
A is a letter if capitalized and not the first word in the sentence
I, A, and a are considered words in all other cases

Note: Periods are removed from the acronyms.
Note on capitalization
Vocalizers sophisticated handling of capitalized characters can affect
interpretation of input if words are capitalized incorrectly. When sending proper
names or other words through the system, make sure they are sent to Vocalizer
the way they would appear normally, with appropriate capitalization, rather
than using all capital letters.
TIP: When you know in advance that text input includes the names of places,
make sure you capitalize them appropriately. Sending New York instead of
new york gives Vocalizer a better chance of rendering the name successfully.
This rule also applies to your custom text replacement filters (see Creating
custom filters for text replacement on page 52).
1M one million
1m one meter
5yt five y t
Numbers
77
Numbers
Vocalizer has a set of defined conventions for handling digits and common
components that contain numbers such as dollar amounts and telephone
numbers. Use the rules defined in this section to ensure the most accurate and
natural synthesis for specific types of numbers and number-based phrases.
Cardinals and ordinals
Vocalizer converts cardinal numbers (which indicate quantity) and ordinal
numbers (which indicate position in a series) to their textual form.
Using a comma within the number does not change speech output:
However, if the comma is followed by one or more spaces, Vocalizer treats the
digits as separate numbers:
Digit sequences
Digit sequences are spoken in groups based on the way you transcribe them. To
have digits voiced individually instead of as a single number, include either:
A space between digits
A comma followed by a space
A period followed by a space

These examples show how various digit strings are converted to speech:
56,734 fifty six thousand seven hundred and thirty four
56734 fifty six thousand seven hundred and thirty four
1st first
345th three hundred and forty fifth
1,103rd one thousand one hundred and third
56, 734 fifty six (pause) seven hundred and thirty four
1, 103rd one (pause) one hundred and third
Developers Guide
78

Periods in digit
sequences
When periods (not followed by a space) are used as separators in a digit
sequence, Vocalizer interprets the string as either a date, if the sequence makes
sense as a date, or an IP address:

For more information, see Dates on page 84.
Leading zeros Leading zeros in a number are omitted from the speech output:
Note: The digit 0 is always said as zero.
1234567 one million two hundred thirty four thousand five
hundred sixty seven
1 2 3 4 5 6 7 one two three four five six seven
(No pauses)
1, 2, 3, 4, 5, 6, 7 one two three four five six seven
(Short pauses)
1. 2. 3. 4. 5. 6. 7 one two three four five six seven
(Long pauses)
Input text Output: North American format Output: U.K. and Australian/New
Zealand format
1.20.2000 the twentieth of january two thousand one point two zero point two zero zero
zero
1.32.1999 one point three two point one nine nine
nine
one point three two point one nine nine
nine
13.1.2001 thirteen point one point two zero zero one the thirteenth of january two thousand
and one
0123 one hundred and twenty three
01 one
Numbers
79
Fractions
Vocalizer recognizes fractions, where numerators and denominators are
separated by a forward slash, with no intervening spaces:
Decimal fractions
Vocalizer reads decimal numbers naturally, incorporating the decimal point in
the phrase.
To ensure decimal values between 0 and 1 are read correctly, you must use a
leading zero. There is no limit to the number of permissible decimal places. For
example:
If the input text begins with a decimal point and includes letter characters, the
decimal point is read as dot and the digits are read as a single number. If the
decimal point is internal, it is omitted from the output. For example:
1/2 one half
3/4 three quarters
42/357 forty two three hundred and fifty sevenths
1 / 2 one slash two
24,563.75 twenty four thousand five hundred and sixty three point
seven five
3.141 three point one four one
0.76 zero point seven six
.50 dot fifty
.9k dot nine k
p.50 p fifty
Developers Guide
80

Percentages
The percent sign (%) is always read as percent. Space between the number
and the percentage sign does not matter. For example:
Equations
Vocalizer correctly reads equations containing arithmetic operators such as plus
(+), minus (-), and equals (=):
Account and social security numbers
To have account numbers or social security numbers read in segments instead of
as a single digit, you must include hyphens between the segments:
Combining letters and numbers
Combining letters and numbers in a single words generates varied results.
Typically the characters are read individually and the digit groups are treated as
a single number:
75% seventy five percent
75 % seventy five percent
99.9% ninety nine point nine percent
0.99% zero point nine nine percent
3+4 three plus four
18-4=14 eighteen minus four equals fourteen
555-00-9800 five five five zero zero nine eight zero zero
5405-9870-8000 five four zero five nine eight seven zero eight zero zero
zero
h1 h one
Currency
81
Currency
Vocalizer recognizes the following currency symbols:
Dollar sign ($)
Pound sterling sign ()
Euro symbol ()
French franc abbreviation (FF)
Deutschmark abbreviation (DM)

Numbers preceded by a recognized currency symbol (possibly with intervening
spaces) are interpreted as currency amounts and expanded appropriately, as
shown:
hello123 h e l l o one two three
45z45 four five z four five
23b twenty three b
$12.34 twelve dollars and thirty four cents
$ 12.34 twelve dollars and thirty four cents
12.34 twelve pounds and thirty four pence
12.34 twelve pounds and thirty four pence
12.34 twelve euros and thirty four cents
12.34 twelve euros and thirty four cents
FF12.34 twelve francs and thirty four centimes
FF 12.34 twelve francs and thirty four centimes
DM12.34 twelve deutschmarks and thirty four pfennigs
DM 12.34 twelve deutschmarks and thirty four pfennigs
$1599.99 one thousand five hundred and ninety nine dollars and
ninety nine cents
Developers Guide
82

Telephone numbers
Vocalizer recognizes local, national, and international telephone number
formats used in North America and the United Kingdom, including area codes,
access prefixes, and the international indicator prefix (+). If a three-digit
segment of the number ends with exactly two zeros, it is read as hundred.
The digits are grouped as they are presented, in order that pauses may be
introduced:
$2 million two million dollars
$1000.896 one thousand dollars and eighty nine point six cents
$0.99 ninety nine cents
.99 ninety nine pence
-$.99 minus ninety nine cents
-$1.95 minus one dollar and ninety five cents
North American formats
847-5900 eight four seven (pause) five nine zero zero
847 9999 eight four seven (pause) nine nine nine nine
800 9999 eight hundred (pause) nine nine nine nine
650-847-9999 six five zero (pause) eight four seven (pause) nine nine nine nine
650.847.9999 six five zero (pause) eight four seven (pause) nine nine nine nine
[650]847.9999 six five zero (pause) eight four seven (pause) nine nine nine nine
(800)555-0202 eight hundred (pause) five five five (pause) zero two zero two
1-800-555-1212 one eight hundred (pause) five five five (pause) one two one two
1 888 847 9999 one eight eight eight (pause) eight four seven (pause) nine nine nine nine
1888 847 9999 one eight eight eight (pause) eight four seven (pause) nine nine nine nine
+1 888 847 9999 plus one eight eight eight (pause) eight four seven (pause) nine nine nine nine
Addresses
83
Addresses
Addresses are handled as follows:
Zip codes
A five-digit number, optionally followed by a hyphen and a four-digit number,
is treated as a zip code if it is immediately preceded by the two-character postal
U.K. formats
456 161 four five six (pause) one six one
456161 four five six one six one
222 1234 two two two (pause) one two three four
8222 1234 eight two two two (pause) one two three four
020 8222 1234 zero two zero (pause) eight two two two (pause) one two three four
(020) 8222 1234 zero two zero (pause) eight two two two (pause) one two three four
01603 456 161 zero one six zero three (pause) four five six (pause) one six one
(01603) 456 161 zero one six zero three (pause) four five six (pause) one six one
+44 20 8222 1234 plus four four (pause) two zero (pause) eight two two two (pause) one two three four
Australian/New Zealand formats
9818 7882 nine eight one eight (pause) seven eight eight two
02 9818 7882 zero two (pause) nine eight one eight (pause) seven eight eight two
0407 060 757 zero four zero seven (pause) zero six zero (pause) seven five seven
+61 2 9818 7882 plus six one (pause) two (pause) nine eight one eight (pause) seven eight eight two
+61 407 060 757 plus six one (pause) four zero seven (pause) zero six zero (pause) seven five seven
1380 Willow Road thirteen eighty willow road
180 Park Avenue one hundred and eighty park avenue
42 St Jamess St forty two saint jamess street
Developers Guide
84

abbreviation of a state name. The numbers in a zip code are always read
individually. For example:
Note: Vocalizer handles Canadian postal codes and UK post codes as described
in Combining letters and numbers on page 80.
Dates
Vocalizer recognizes dates specified in both text (December 9, 1997) and digit
(12/9/1997) formats, and reads them accordingly. The digit format can be
delimited with forward slashes (/), commas(,), periods (.), or hyphens
(-).
Note: If you are using North American English voice packs, dates are parsed
using the American date format, mm/dd/yyyy. If you are using either U.K. or
Australian/New Zealand English voice packs, dates are parsed using the
dd/mm/yyyy format.
Vocalizer recognizes month abbreviations as well as the full month name.
Certain month names and month abbreviationssuch as Jan, Mar, May, and
Juneare treated specially because they are also words. Based on the context of
the overall phrase, these might be read as dates or read literally:
94025 ninety four thousand and twenty five
CA 94025 california nine four zero two five
Menlo Park, CA 94025 menlo park california nine four zero two five
Florham Park, NJ
07932-0971
florham park new jersey
zero seven nine three two zero nine seven one
Menlo Park, 94025 menlo park ninety four thousand and twenty five
California 94025 california ninety four thousand and twenty five
Zealand format
Feb 5 february five february five
Jan 1, 2000 january one two thousand january one two thousand
February 5, 1997 february five nineteen ninety seven february five nineteen ninety seven
Times
85
Times
Vocalizer recognizes time-of-day phrases in the formats hh:mm and hh:mm:ss,
where the hour is between 0 and 24. It also recognizes time phrases that are
succeeded by the following abbreviations: am, pm, AM, or PM. Space between
the number and the abbreviation does not matter.
Time-of-day phrases in the format X:00 are converted to X oclock. Hours from
13 to 24 are read using military time. If the hour value is greater than 24, the
phrase is not read as a time. For example:
1/1/2000 the first of january two thousand the first of january two thousand
3/12/97 the twelfth of march ninety seven the third of december ninety seven
3.12.97 the twelfth of march ninety seven the third of december ninety seven
1/6/99 the sixth of january ninety nine the first of june ninety nine
1/6/00 the sixth of january two thousand the first of june two thousand
1/6/01 the sixth of january two thousand and one the first of june two thousand and one
3/12 three twelfths three twelfths
on 3/12 on three twelfths on three twelfths
12/1 twelve fslash one twelve slash one
4-19-1966 the nineteenth of april nineteen sixty six four one nine one nine six six
4-19 four minus nineteen four minus nineteen
19-4 nineteen minus four nineteen minus four
Jan 1 january one january one
1 Jan one january one january
Zealand format
1:30 one thirty
1:30:31 one thirty and thirty one seconds
6:30 am six thirty am
Developers Guide
86

Email and web addresses
Vocalizer reads email and web addresses naturally, including symbols such as
@ (expanded to at) and . (expanded to dot). Vocalizer determines when
address segments can be read as words or need to be spelled out. For example:
Handling 8-bit characters
If you expect to enter foreign language words or expressions in ssml mode, you
must specify the encoding as described in Encoding on page 39.
6:30am six thirty am
12:00 twelve oclock
12:00:05 twelve oclock and five seconds
05:45 five forty five
23:00 twenty three hundred hours
14:35 fourteen thirty five
34:34 thirty four thirty four
[email protected] mail at nuance dot com
[email protected] a b c at nuance dot com
www.nuance.com w w w dot nuance dot com
http://www.nuance.com h t t p colon slash slash w w w dot nuance dot com
http://extranet.nuance.com
/developers/index.html
h t t p colon slash slash extranet dot nuance dot com
slash developers slash index dot h t m l
Appendix A Text processing for Canadian French
A
87
Appendix A
Text processing for
Canadian French
This appendix outlines how Nuance Vocalizer processes text input for Canadian
French.
Topics include:
Text encoding
Abbreviations
Capitalization
Numbers
Currency
Telephone numbers
Addresses
Dates
Times

For detailed information on text normalization rules used for English dialects,
see Chapter 9.
Developers Guide
88

Text encoding
Nuance Vocalizer supports the Microsoft CP1252 (WinLatin1) and ISO-8859-1 If
you are using Vocalizer in SSML mode, Vocalizer can interpret ISO-8859-1 input
characters if you provide an appropriate XML header:
<speak> Les lettres accentues telles que: ,,,...sont
frquentes en franais </speak>
For more information, see Encoding on page 39.
paragraphs. Vocalizer strips the layout from input text and uses it in conjunction
with punctuation to determine phrasing, thereby enabling natural-sounding
into multiple sentences.
This section outlines text normalization rules for:
Periods
Hyphens and dashes

For general text normalization rules used to interpret punctuation, see the
Phrasing and punctuation on page 69.
Periods
Vocalizer determines the correct interpretation for periods from the context.
Periods can be used as:
Sentence terminatorsAs a general rule, allow two spaces to follow a

sentence terminator
Abbreviation markersSee Abbreviations on page 89
DecimalsSee Decimal fractions on page 94

Abbreviations
89
Hyphens and dashes
Since hyphens, dashes, and minus signs are represented by the same ASCII
character, ambiguous situations may occur at times. To avoid ambiguous
situations, Vocalizer uses a set of disambiguation rules. Specifically, Vocalizer
considers the - character to be:
A trait dunion if located between letters and consider each side as being part
of the same word if in dictionary.
A signe de moins if followed by a number, indicating a negative number.
A tiret and pronounced if located between two numbers without a space

as in number intervals or game scores.
A trait dunion if preceded by a letter and followed by a carriage return plus a

letter. The words on either side of the hyphen are joined and handled as a
single word.
For example:
Abbreviations

Input text Output
(salut) salut
salut) salut parenthse
salut salut
salut salut
Developers Guide
90

Single-letter abbreviations, homographic abbreviations, and measurement
abbreviations are described next.
Note: As a general rule, use a period and an optional space to terminate an
abbreviation.
the letter as:

of abbreviations


meaning, for example, the noun lead and the verb lead. Homographic
abbreviations are abbreviations that are spelled the same way as a word or letter
but have different meaning, for example, the abbreviation for jeudi (jeu.) and
jeu the noun.
homographic abbreviations. For example, a filter that replaces on with
Ontario could incorrectly render the word on when encountered in regular
text.
Input text Output
jeu. 4 dcembre jeudi quatre dcembre
jeu.. 4 dcembre jeudi (pause) quatre dcembre.
Va au jeu. Alors, nous
partirons.
Va au jeu (pause) alors
Capitalization
91
digit (with or without a space). For example:
initials as follows so that the pronunciation tool can treat them appropriately:
first letter)
A single letter, except where the letter is an abbreviation or A, a, ,

, Y, and y
See Single-letter abbreviations on page 90.
Capitalization
than using all or no capital letters.
Input text Output
5m cinq mtres
+6.6cm plus six point six centimtres
7 km sept kilomtres
-40kg moins quarante kilogrammes
250ml deux cent cinquante millilitres
23s vingt trois secondes
30 C trente degr Celsius
-15 degr F moins quinze degr Fahrenheit
Developers Guide
92

This rule also applies to your custom text replacement filters, which are
discussed in the Creating custom filters for text replacement on page 52.
Directions
Use appropriate capitalization to prevent ambiguity when text input includes
directions. For example, the verb est and the direction Est would be
rendered incorrectly in the following text:
La circulation est congestionne en direction est sur le boulevard Maisoneuve
est.
Numbers
Cardinals and ordinals
Vocalizer converts cardinal numbers (which indicate quantity) and ordinal
numbers (which indicate position in a series) to their textual form.
The metric system is the main convention supported, but the imperial system is
also supported when there is no conflict with the metric system. The metric
convention uses single spaces to separate digit triplets within integers (which
indicate hundreds, thousands, millions, and so on). While Vocalizer accepts
commas as separators for digit triplets (imperial system), Nuance recommends
using spaces instead to avoid confusion with commas as decimal separators.
Input text Output
5600 cinq mille six cents
5 600 cinq mille six cents
4500000 quatre millions cinq cents mille
4ime quatrime
102e cent deuxime
Numbers
93
However, if the comma is followed by one or more spaces, Vocalizer treats the
digits as separate numbers:
Digit sequences

While a single space or a comma with no space can be used to enumerate, they
are not recommended as they could be interpreted as triplet separators within a
single number.
These examples show how to various digit strings are converted to speech:
1234ime mille deux cent trente-quatrime
Input text Output
56, 734 cinquante-six (pause) sept cent trente-quatre
1, 103ime un (pause) cent troisime
Input text Output
Input text Output
1234567 un million deux cent trente-quatre mille cinq cent
soixante-sept
4, 500, 200 quatre (pause) cinq cents (pause) deux cents
1, 2, 3, 4, 5, 6, 7 un deux trois quatre cinq six sept
(Short pauses)
1. 2. 3. 4. 5. 6. 7 un deux trois quatre cinq six sept
(Long pauses)
4 5 6 7 quatre cinq six sept
4 500 200 quatre millions cinq cents mille deux cents
Note: This example illustrates the disadvantage of using
single spacing as a separator, since the digits are
processed as a single number.
Developers Guide
94

Fractions
Decimal fractions
By default, Vocalizer interprets metric decimal number convention for Canadian
French, which uses a comma to separate decimals from the integer part.
Optionally, spaces could be used to separate the digit triplets (hundreds,
thousands, millions, and so on). Usually, the decimal part is a digit enumeration,
except for digit pairs higher than 09. For example:
The imperial decimal system, using a period to separate the integer from the
decimal part, is also supported when there is no possible ambiguity. For
example, the number 1,234 would be considered by default as a metric decimal
number and would be synthesized as un virgule deux trois quatre. However,
combining the use of a comma separator and a period, (1,234.5) forces the
Input text Output
044 quarante-quatre
02 deux
00 zro
Input text Output
1/2 une demie
-13/5 moins treize cinquime
1 / 2 un barre oblique deux
Input text Output
+1300321 plus un million trois cents mille trois cent vingt et un
4,15 quatre virgule quinze
4.66 quatre point soixante-six
-55,345 moins cinquante-cinq virgule trois quatre cinq
6,05 six virgule zro cinq
Numbers
95
number to be recognized as an imperial decimal number, and the output would
be mille deux cent trente quatre point cinq. For example:
Percentages
The percent sign (%) is always read as pour cent. Space between the
number and the percentage sign does not matter. For example:
Account numbers
as a single digit, you must include hyphens or slashes between the segments.
The character separator is pronounced to help determine segment boundaries:
Input text Output
13,400,777 treize millions quatre cents mille sept cents
-4.72 moins quatre point soixante-douze
56,400.924 cinquante-six mille quatre cents point neuf deux quatre
56,466 cinquante-six virgule quatre six six
Note: Parsed by default as a metric decimal.
Input text Output
90% quatre-vingt-dix pour cent
-5.125 % moins 5 point un deux cinq pour cent
+700 % plus sept cents pour cent
9/4% neuf quart pour cent
Input text Output
2222/3333/4444/5555 deux deux deux deux barre oblique trois trois trois trois
barre oblique quatre quatre quatre quatre barre oblique
cinq cinq cinq cinq
44/33 quarante-quatre trente-troisieme
Note: This example is interpreted as a fraction rather
than an account number because there are only two
numbers separated by a slash with no intervening
spaces. See Fractions on page 94.
Developers Guide
96

Typically the characters are read individually and digits groups are treated as a
single number if at the beginning:
78-122-2 sept huit trait dunion un deux deux trait dunion deux
78-122 soixante-dix-huit cent vingt-deux
Note: This example is interpreted as an interval or a
game score rather than an account number.
Input text Output
Input text Output
c20b c deux zero b
d4df3 d quatre d f trois
20b vingt b
Currency
97
Currency
Vocalizer recognizes the following currency symbols:
Dollar sign ($), including:
Canadian dollars ($CAD)
American dollars ($USD)
Cent sign ()
Pound sterling sign ()
Euro symbol ()
French franc abbreviation (FF)
Deutschmark abbreviation (DM)
Yen symbol ()
Vocalizer supports the official French convention of currency symbols following
the digits and will also accept currency symbols preceding the digits, as in
English, which allows the abbreviation for millions or billions suffix to be
added.
Input text Output
50$ cinquante dollars
12$ CAD douze dollars canadien
12 $USD douze dollars amricain
$25.50 vingt-cinq dollars et cinquante cennes
10,40 $ CAD dix dollars canadiens et quarante cennes
$0,99 quatre-vingt-dix-neuf cennes
65,66USD soixante-cinq virgule soixante-six cennes amricain
$50M cinquante millions de dollars
10.20 dix livres et vingt penny
12,34 douze virgule trente-quatre euros
30,75FF trente francs et soixante-quinze centimes
12,34 DM douze virgule trente-quatre deutsche mark
Developers Guide
98

Telephone numbers
Vocalizer recognizes local and national telephone number formats used in North
America, including area codes.
introduced:
Addresses
Vocalizer recognizes zip and postal codes and address abbreviations, such as
ave or st. Addresses are handled as follows:
100 cent yen
Input text Output
Input text Output
256 9866 deux cinq six (pause) neuf huit six six
514-732-4619 cinq un quatre (pause) sept trois deux (pause)
quatre six un neuf
(819) 623-4455 huit un neuf (pause) six deux trois (pause)
quatre quatre cinq cinq
1 800 256 9866 un (pause) huit cents (pause) deux cinq six (pause)
neuf huit six six
1-514-732-4619 un (pause) cinq un quatre (pause) sept trois deux
(pause) quatre six un neuf
Input text Output
Alexandre LeGrand
101 3e ave.
Mtl, Qc
H5T 2v2
Alexandre LeGrand cent et un troisieme avenue (pause)
Montral (pause) Qubec h cinq t deux v deux
111 Duke, Mtl, Qc,
Canada, H8T2V7
cent onze Duke (pause) Montral (pause) Qubec (pause)
Canada (pause) h huit t deux v sept
CA 22333 Californie deux deux trois trois trois
Dates
99
Note: Pauses are added when there is punctuation between address fields.
Dates
Vocalizer recognizes dates specified in both text (9 dcembre, 2003) and digit
(2003/12/09) formats, and reads them accordingly. The digit format can be
delimited with forward slashes (/), periods (.), or hyphens (-). Normally,
the digit order is year first, followed by month then day, but the reverse order is
also supported (day/month/year) when no ambiguity is possible with the main
convention. Standard abbreviations for months are supported.
CO. 44666-1234 Colorado quatre quatre six six six trait dunion un deux trois
quatre
Input text Output
Input text Output
1999/1/5 le cinq janvier mille neuf cent quatre-vingt-dix-neuf
13-12-2003 le treize dcembre deux mille trois
99/03/04 le quatre mars mille neuf cent quatre-vingt-dix-neuf
01/02/03 le trois fvrier deux mille un
02/4/5 le cinq avril deux mille deux
6 jan, 2002 le six janvier deux mille deux
4 fev le quatre fvrier
2003-04-02 le quatre fvrier deux mille trois
14/14/2000 un quatre barre oblique un quatre barre oblique deux zro
zro zro
Note: This example can not be a valid date so it is
interpreted as an account number. See Account numbers
on page 95.
Developers Guide
100

Times
Vocalizer recognizes time-of-day phrases using h and min to delimit the
hour and minute fields. It also recognizes time phrases that are succeeded by the
abbreviations am, pm, AM, or PM with a space between the number
and the abbreviation.
@ (expanded to a commercial) and . (expanded to point). Vocalizer
determines when address segments can be read as words or need to be spelled
out. For example:
Input text Output
5h30min30sec cinq heures trente minutes et trente secondes
13h45 treize heures quarante-cinq minutes
4 h 20 min quatre heures vingt minutes
20h vingt heures
20h00 vingt heures
0h30 zro heures trente minutes
6h pm six heures p m
3h 30min am trois heures trente minutes a m
9h AM neuf heures a m
Input text Output
[email protected] courrier a commercial nuance point com
[email protected] a b c a commercial nuance point com
www.nuance.com triple w point nuance point com
http://www.nuance.com h t t p deux points barre oblique barre oblique triple w
point nuance point com
101
http://support.nuance.com
/developers/index.html
h t t p deux points barre oblique barre oblique support
point nuance point com barre oblique developers barre
oblique index point h t m l
Input text Output
Developers Guide
102

Appendix B Text processing for American Spanish
B
103
Appendix B
Text processing for
American Spanish
This appendix outlines how Nuance Vocalizer processes text input for
American Spanish.
Topics include:
Text encoding
Regionalism
Abbreviations
Capitalization
Numbers
Currency
Telephone numbers
Dates
Times
Addresses

For detailed information on text normalization rules used for English dialects,
see Chapter 9.
Developers Guide
104

Text encoding
Nuance Vocalizer supports the Microsoft CP1252 (WinLatin1) and ISO-8859-1
character encoding for plaintext input of American Spanish characters.
If you are using Vocalizer in SSML mode, Vocalizer can interpret ISO-8859-1
input characters if you provide an appropriate XML header:
<speak> Las letras acentuadas son muy usadas en espaol </speak>
For more information, see Encoding on page 39.
Regionalism
For American Spanish, Vocalizer supports various conventions for numbers,
dates, and addresses. For example, in the United States, decimal numbers use
the form 4,123.5, while most Central and Latin American countries use the form
of 4.123,5. Since these variations cause conflicting interpretations, Vocalizer
supports the -region command-line option to determine the normalization
convention. Set -region to either US or CALA to specify the set of conventions
required. See Using the -region option on page 12.
The main distinction between CALA and US mode are:
paragraphs. Vocalizer strips the layout from input text and uses it in conjunction
with punctuation to determine phrasing, thereby enabling natural-sounding
US CALA
Periods indicate decimal points; digit
triplets are separated with commas
Commas indicate decimal points; digit
triplets are separated with periods
Dates are interpreted using the
mm/dd/yyyy format
Dates are interpreted using the
dd/mm/yyyy format
$ indicates dollars $ indicates pesos
Address abbreviations are
interpreted in English
Address abbreviations are
interpreted in Spanish
105
into multiple sentences.
This section outlines text normalization rules for:
Periods

For general text normalization rules used to interpret punctuation, see the
Phrasing and punctuation on page 69.
Periods
Vocalizer determines the correct interpretation for periods from the context.
Periods can be used as:
Sentence terminatorsAs a general rule, allow two spaces to follow a

sentence terminator
Abbreviation markersSee Abbreviations on page 106
DecimalsSee Cardinals and decimal fractions on page 108

For example:
Input text Output
(buenos das) buenos das
buenos das) buenos das cierra parntesis
buenos das buenos das
buenos das buenos das
Developers Guide
106

Abbreviations

Single-letter abbreviations, homographic abbreviations, and measurement
abbreviations are described next.
Note: As a general rule, use a period and an optional space to terminate an
abbreviation.
the letter as:

of abbreviations


Abbreviations
107
meaning, for example, the adjective dedicada and the verb dedicada.
Homographic abbreviations are abbreviations that are spelled the same way as a
word or letter but have different meaning, for example, the abbreviation for
marzo (mar.) and mar the noun.
homographic abbreviations. For example, a filter that replaces mar with
marzo could incorrectly render the word mar when encountered in regular
text.
digit (with or without a space). For example:
Input text Output
4 mar. 2004 cuatro de marzo de dos mil cuatro
Vivo cerca del mar. Vivo cerca del mar.
Input text Output
5m cinco metros
+6.6cm ms seis punto seis centmetros
7 km siete kilmetros
-40kg menos cuarenta kilogramos
250ml doscientos cincuenta mililitros
23s veintitrs segundos
30 C treinta grados centgrados
-15 grados F menos quince grados Fahrenheit
Developers Guide
108

initials as follows so that the pronunciation tool can treat them appropriately:
first letter)
A single letter, except where the letter is an abbreviation or A, a, E,

e, O, o, U, u, Y, and y
See Single-letter abbreviations on page 106.
Capitalization
than using all or no capital letters.
This rule also applies to your custom text replacement filters, which are
discussed in the Creating custom filters for text replacement on page 52.
Numbers
Cardinals and decimal fractions
Vocalizers interpretation of cardinal numbers depends on the value of the
-region option (see Regionalism on page 104). In US mode, interpretation of the
text input is based on the English system, which uses commas to separate digit
triplets (indicating hundreds, thousands, millions, and so on) and periods
Numbers
109
indicate decimals. In CALA mode, periods are used to separate digit triplets and
commas indicate decimals. When there is no ambiguity, the appropriate
convention is used regardless of the -region setting.
For example:
If the comma is followed by one or more spaces, Vocalizer treats the digits as
separate numbers:
Ordinals
Vocalizer interprets the two frequent forms of Spanish ordinals: the simo
form and the avo form. The ordinal suffix determines the form, gender, and
number associated with the ordinal. You can use suffixes with 1 to 4 letters.
Input text US mode CALA mode
5601 cinco mil seiscientos uno cinco mil seiscientos uno
5,601 cinco mil seiscientos uno cinco coma seis cero uno
5.601 cinco punto seis cero uno cinco mil seiscientos uno
-4.500,05 menos cuatro mil quinientos
coma cero cinco
menos cuatro mil quinientos
coma cero cinco
4.12 cuatro punto doce cuatro punto doce
4.123 cuatro punto uno dos tres cuatro mil ciento veintitrs
Input text Output
56, 734 cincuenta y seis, setecientos treinta y cuatro
Input text Output
1o primero
1os primeros
1ro primero
1ero primero
1eros primeros
1a primera
7o sptimo
Developers Guide
110

Digit sequences

While a single space or a comma with no space can be used to enumerate, they
are not recommended as they could be interpreted as triplet separators within a
single number.
These examples show how to various digit strings are converted to speech:
7os sptimos
7mo sptimo
7imo sptimo
7imos sptimos
3o tercero
5as quintas
17o decimosptimo
32a trigsima segunda
32avo treinta y dosavo
17avo diecisieteavo
17ava diecisieteava
17avos diecisieteavos
17avas diecisieteavas
Input text Output
Input text Output
1234567 un milln doscientos treinta y cuatro mil quinientos
sesenta y siete
4, 500, 200 cuatro (pause) quinientos (pause) doscientos
Numbers
111
Fractions
Percentages
The percent sign (%) is always read as por ciento. Space between the
number and the percentage sign does not matter. For example:
1, 2, 3, 4, 5, 6, 7 uno dos tres cuatro cinco seis siete
(Short pauses)
1. 2. 3. 4. 5. 6. 7 uno dos tres cuatro cinco seis siete
(Long pauses)
4 5 6 7 cuatro cinco seis siete
4,500,200 cuatro millones quinientos mil doscientos (US mode)
Note: This example illustrates the disadvantage of using
single coma as a separator, since the digits are
processed as a single number.
Input text Output
Input text Output
044 cuarenta y cuatro
02 dos
00 cero
Input text Output
1/2 un medio
-13/5 menos trece quintos
Input text Output
90% noventa por ciento
-5.125 % menos cinco mil ciento veinticinco por ciento
+700 % ms setecientos por ciento
Developers Guide
112

Account numbers
as a single digit, you must include hyphens or slashes between the segments.
The character separator is pronounced to help determine segment boundaries:
Typically the characters are read individually and digits groups are treated as a
single number if at the beginning:
Currency
Vocalizer recognizes currency symbols, including:
9/4% nueve cuartos por ciento
Input text Output
Input text Output
2222/3333/4444/5555 dos dos dos dos barra oblicua tres tres tres tres barra
oblicua cuatro cuatro cuatro cuatro barra oblicua cinco
cinco cinco cinco
44/33 cuarenta y cuatro treinta y tresavos
Note: This example is interpreted as a fraction rather
than an account number because there are only two
numbers separated by a slash with no intervening
spaces. See Fractions on page 111.
78-122-2 siete ocho guin uno dos dos guin dos
78-122 setenta y ocho (pause) a (pause) ciento veintids
Note: This example is interpreted as an interval or a
game score rather than an account number.
Input text Output
c20b c dos cero b
d4df3 d cuatro d f tres
20b veinte b
Telephone numbers
113
Dollar sign ($, $US, US$), including:

Note: The interpretation of $ depends of the region specified (see
regionalism above section) being dlares for US mode and pesos for
CALA mode.
Cent sign ()
Pesos symbols ($, P, Ps, PS)
Nuevo pesos (N$, N$U)
ISO currency symbols, such as MXN, CLP, ARS, USD, and so on

Vocalizer accepts currency symbols either preceding or following digits, and
abbreviations for million or billion suffix.
Telephone numbers
Vocalizer recognizes local and national telephone number formats used in North
America, including area codes.
introduced:
Input text Output
$50 US mode: cincuenta dlares
CALA mode: cincuenta pesos
US $22M veintids millones de dlares
$25.50 veinticinco dlares con cincuenta centavos
$0,99 noventa y nueve centavos
21 P veintin pesos
101 MXN ciento un pesos mexicanos
Input text Output
256 9866 dos cinco seis (pause) nueve ocho seis seis
514-732-4619 cinco uno cuatro (pause) siete tres dos (pause) cuatro
seis uno nueve
(232) 623-4455 dos tres dos (pause) seis dos tres (pause) cuatro
cuatro cinco cinco
Developers Guide
114

Dates
Vocalizer recognizes dates specified with abbreviation (6 ene. 2004) or digit
format (2004/01/06).
The expected digit format depends on the -region specified (see Regionalism
on page 104). In US mode, date input is interpreted using the format mm/dd/yyyy
(or mm/dd/yy). In CALA mode, date input is interpreted using the format
dd/mm/yyyy (or dd/mm/yy). You can delimit digit formats with forward slashes
(/), hyphens (-) or periods (.).
When date input is not entered in the expected order, alternative forms are
considered. For example, if Vocalizer is in CALA mode, Vocalizer recognizes
that a date such as 02/30/04 cannot be processed as yy/mm/dd but should be
interpreted as mm/dd/yy.
1 800 256 9866 uno (pause) ochocientos (pause) dos cinco seis
(pause) nueve ocho seis seis
2335-4445 dos tres tres cinco (pause) cuatro cuatro cuatro cinco
233-44-45 dos tres tres (pause) cuatro cuatro (pause) cuatro
cinco
(23)(23) 3453456 dos tres (pause) dos tres (pause) tres cuatro cinco tres
cuatro cinco seis
Tel: 623-4455 telfono (pause) seis dos tres (pause) cuatro cuatro
cinco cinco
TEL 2334545 telfono (pause) dos tres tres cuatro cinco cuatro cinco
Input text Output
Input text Output
1999/1/5 cinco de enero de mil novecientos noventa y nueve
13-12-2003 trece de diciembre de dos mil tres
99/03/04 cuatro de marzo de mil novecientos noventa y nueve
01/02/03 US mode: dos de enero de dos mil tres
CALA mode: primero de febrero de dos mil tres
02/4/5 US mode: cuatro de febrero de dos mil cinco
CALA mode: dos de abril de dos mil cinco
Times
115
Times
Vocalizer recognizes time-of-day phrases using h and min to delimit the
hour and minute fields or the semicolon form hh:mm:sec. It also recognizes time
phrases that are succeeded by the abbreviations am, pm, GMT, or EST
with a space between the number and the abbreviation.
6 Ene. 2004 seis de enero de dos mil cuatro
12 Ago., 2000 doce de agosto de dos mil
2003-04-02 dos de abril de dos mil tres
14/14/2000 uno cuatro barra oblicua uno cuatro barra oblicua dos cero
cero cero
Note: This example can not be a valid date so it is
interpreted as an account number. See Account numbers
on page 112.
Input text Output
Input text Output
5h30min30sec cinco horas treinta minutos y treinta segundos
13h45 trece horas y cuarenta y cinco
4 h 20 min cuatro horas y veinte minutos
20h veinte horas
20h00 veinte horas
0h30 cero hora y treinta
6h pm seis horas postmeridiano
3h 30min am tres horas y treinta minutos antemeridiano
9h AM nueve horas antemeridiano
5 :30 :20 cinco horas treinta minutos y veinte segundos
4 :20 cuatro horas y veinte
12 :00 medioda
8h GMT ocho horas (pause) hora media de Greenwich
Developers Guide
116

Addresses
Vocalizer recognizes address elements like country abbreviations, state
abbreviations, postal codes, and address abbreviations, such as ave., st,
blvd., N., Nte, and P.O.
Typical abbreviations for both English and Spanish address formats are
supported. Interpretation of some abbreviations depends on the -region setting
specified (see Regionalism on page 104). For example N. is translated as
North in US mode but as Norte in CALA mode.
To facilitate the interpretation of the address rendering, short pauses are
introduced to separate the most important element of the address. Addresses
are handled as follows:
9h30 EST nueve horas y treinta (pause) hora estndar del este
10:40 PST diez horas y cuarenta (pause) hora estndar del
pacfico
Input text Output
Input text Output
Miami, FL Miami (pause) Florida
FL, USA Florida (pause) Estados Unidos
232 El Camino Real Blvd.
Menlo Park, CA
doscientos treinta y dos (pause) El Camino Real
boulevard (pause) Menlo Park (pause) California
444 25th Ave
Santa Claus, CA
cuatrocientos cuarenta y cuatro (pause) twenty fifth
Avenue (pause) Santa Claus (pause) California
12 N Mormon St.
Salt Lake City
UT 12345-6666 USA
US mode: Doce (pause) North Mormon Street (pause)
CALA mode: Doce (pause) Norte Mormon Street (pause)
Salt Lake City (pause) Utah (pause) uno dos tres cuatro
cinco (pause) seis seis seis (pause) Estados Unidos
40 E Cactus Dr.
Suite 20c
Tucson, AZ
Cuarenta (pause) East Cactus Drive (pause) suite viente
c (pause) Tucson (pause) Arizona
1000 Redwood
PO BOX 101
Seattle, 90001
USA
Mil (pause) Redwood (pause) Post Office Box ciento uno
(pause) Seattle (pause) nueve cero cero cero uno
(pause) Estados Unidos
Addresses
117
Mrs. Chandler
7449 Rupert Avenue
SAINT LOUIS, MO
USA 63117
Misses Chandler (pause) siete mil cuatrocientos cuarenta
y nueve (pause) Rupert Avenue (pause) Saint Louis
(pause) Missouri (pause) Estados Unidos (pause) seis
tres uno uno siete
MEX CALA mode: (pause) Mxico
Santiago, CHL CALA mode: Santiago (pause) Chile
Sr. Francisco
Ave Providencia N 112
6 piso, Las Condes
Santiago, CHL 63117
CALA mode: Seor Francisco (pause) avenida
Providencia norte (pause) ciento doce (pause) sexto
(pause) piso (pause) Las Condes (pause) Santiago
(pause) Chile (pause) seis tres uno uno siete
Input text Output
Developers Guide
118

@ (expanded to arroba) and . (expanded to punto). Vocalizer
determines when address segments can be read as words or need to be spelled
out. For example:
Input text Output
[email protected] carlos arroba (pause) nuance punto (pause) com
[email protected] carlos dos b (pause) arroba (pause) nuance punto
(pause) com
[email protected] a (pause) b (pause) c (pause) arroba (pause) nuance
punto (pause) com
www.nuance.com w w w (pause) punto (pause) nuance punto (pause)
com
http://www.nuance.com h t t p (pause) dos punto (pause) barra barra (pause) w
w w (pause) punto (pause) nuance punto (pause) com
Index 119
Index
A
alaw encoding 13
APIs 27
NuanceSpeechChannel 28
RCEngine 29
SpeechObjects 28
audio encoding 13
C
client.TTSAddresses 20
Computer Phonetic Alphabet (CPA) 63
config.LogFileMaxNum 23
config.LogFileMaxSize 23
config.LogfileRoot 23
config.LogFileRootDir 17
D
Dictionary Editor 5, 59-68
adding words 65
audio feedback 62
auto-phonetic transcription 63
deleting words 67
distributing changes 51, 54, 68
features 59
modifying entries 66
port settings 59
testing pronunciation 67
using 64
-dictionary_port 16
digits. See numbers
documentation (Nuance)
accessing xi
description x
E
email parsing 14, 47
Email Reader Configuration Tool 5, 47-51
distributing changes 51
features 47
using 48
-encoding 13, 16
encoding 39
environment variables 7
F
-filter 14, 16
filters
customizing 52
distributing changes 54
order of application 52
specifying multiple filters on
command line 52
using the Text Replacement Filter
Editor 53
G
getting help
Nuance xii
GUI tools
Dictionary Editor 5, 59
Email Reader Configuration Tool 5,
47
Launcher 5
Offline Audio Generator 5, 37
Text Replacement Filter Editor 5, 53
H
-help 16
help
Nuance xii
120 Nuance Vocalizer 3.0
Developers Guide

I
installation 5-7
environment variables 7
Solaris 6
system requirements 3
third-party software requirements 5
Windows 5
J
Java Runtime Environment 4, 48, 59
installing (Windows) 5
upgrading 4
L
Launcher 5, 43-46
configuring Vocalizer 44
starting GUI tools 46
using 43
letters 80, 96, 112
licensing 7
lm.Addresses 17
logging 23
M
markup languages 13
selecting default 13
SSML 13
VoiceXML 13
mulaw encoding 13
N
naming TTS servers 24
Nuance License Manager 7
Nuance System
getting help xii
Nuance Watcher 20-23
NuanceSpeechChannel 28
-num_channels 11, 15
numbers 77, 92, 108
O
Offline Audio Generator 5, 55-57
testing installation 55
testing SSML sample text 37
using 56
validating multi-server
configurations 57
options
-dictionary_port 16
-encoding 13, 16
-filter 14, 16
-help 16
-num_channels 11, 15
-preallocate_ licenses 12, 15
-pruning 15, 16
text formats 13
-text_type 14, 15
-voice 12, 15
P
parameters
client.TTSAddresses 20
config.LogFileMaxNum 23
config.LogFileMaxSize 23
config.LogfileRoot 23
config.LogFileRootDir 17
lm.Addresses 17
rm.Addresses 16
rm.Port 19
tts.Port 17, 25
tts.ResourceName 17
watcher.DaemonStartupFilename
21
watcher.RestartOnFailure 22
PATH 7
phonetic alphabets
Computer Phonetic Alphabet
(CPA) 63
Worldbet 39
XSAMPA 39
Index 121
ports
changing settings 20
dictionary port 25
running two instances on one
machine 25
TTS port 17, 25
-preallocate_ licenses 12, 15
prompt playback functions 27
-pruning 15, 16
punctuation 77, 92, 108
R
RCEngine 29
-region 12, 104, 108, 109, 114, 116
resource manager 17
rm.Addresses 16
rm.Port 19
S
software requirements 5
Solaris
installation 6
SpeechChannel 28
SpeechObjects 28
SSML
elements 32
audio 37
break 33
emphasis 33
paragraph 32
phoneme 32
prosody 35
say-as 33, 34, 35
speak 32
sub 32
encoding 39
sample text 37
URI formats 40
well-formed XML 30
syllable stress marks 63
syntax
Unix xii
Windows xii
T
technical support xii
text formats
expected 14
options 13
text normalization
abbreviations 73, 89, 106
addresses 83, 98, 116
currency 81, 97, 112
dates 84, 99, 114
email addresses 86, 100, 118
numbers
account numbers 80, 95, 112
cardinals 77, 92, 108
combining with letters 80, 96,
112
decimals 79, 94
digit sequences 77, 93, 110
equations 80
fractions 79, 94, 111
leading zeros 78, 94, 111
ordinals 77, 92
percentages 80, 95, 111
telephone numbers 82, 98, 113
periods 71, 88, 105
phrasing and punctuation 69, 88, 104
times 85, 100, 115
web addresses 86, 100, 118
zip codes 83
Text Replacement Filter Editor 5, 53-54
-text_type 14, 15
TTS requests
using SSML 29
tts.Port 17, 25
tts.ResourceName 17, 24
U
URI formats 40
122 Nuance Vocalizer 3.0
Developers Guide

V
vocalizer program 11-17
command-line options 14
connecting to a resource manager 17
Nuance parameters 16
setting tts.ResourceName 24
starting through the watcher 21
-voice 12, 15
voice packs
Claire Kingston 2
installing 8
Josh Donnelly 2
Julie Deschamps 2
Laurie Woods 2
Reed Johnston 2
Sarah Brown 2
selecting 8
Tim Cooper 2
VoiceXML 29
elements 30
audio 31
break 30
div 30
emp 30
pros 31
sayas 31
value 31
encoding 39
URI formats 40
well-formed XML 30
W
Watcher 20-23
watcher.DaemonStartupFilename 21
watcher.RestartOnFailure 22
Windows
installation 5
Worldbet 39
X
XSAMPA 39

Vocalizer 30 Developers Guide PDF

Uploaded by

Copyright:

Available Formats

Vocalizer 30 Developers Guide PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vocalizer 30 Developers Guide PDF

Uploaded by

Copyright:

Available Formats

Nuance Vocalizer

Introduction to the Nuance System, which provides a comprehensive overview

Nuance System Installation Guide, which describes how to install and

Nuance Grammar Developers Guide, which describes the process of designing

Nuance Application Developers Guide, which describes how to develop,

Nuance System Administrators Guide, which describes topics related to

Nuance Platform Integrators Guide, which describes the process and

Nuance API Reference, which includes HTML-based documentation for

Support for Windows 2000 and SPARC Solaris platforms

Dynamic editing of entries and pronunciations in user dictionaries

Support for these voice packs:

With the Nuance System

Standalone (without the Nuance System installed)

For 1-20 channels, 512 MB

For 20-50 channels, 768 MB

On Windows, the JRE version 1.3 is included as part of the installation

Email Reader Configuration Tool

Offline Audio Generator

Text Replacement Filter Editor

Java 2 Runtime Environment (JRE) version 1.3

If you have an earlier version of the JRE, uninstall it before installing

Connect each recognition client directly to a single Vocalizer TTS server

Use a Nuance resource manager to mediate requests between recognition

Specifying the markup language, either plain text, VoiceXML, or SSML

Setting up Vocalizer to accept email files

emailTo enable preprocessing of

Custom filtersTo enable text

The watcher.RestartOnFailure parameter (this tells the Watcher to

Parameters to set up program logging

Starting a Vocalizer TTS server

Quiescing or killing a Vocalizer TTS server

config.LogFileMaxNum controls the number of sequential log files created.

config.LogFileMaxSize sets the maximum size of each log file, in

Invoke a TTS prompt from applications written with Nuance development

Specify a TTS prompt using VoiceXML

VoiceXML 1.0, in accordance with the W3Cs May 5, 2000 Note

8 kHz, mulaw, unheadered

8 kHz, mulaw, Sphere header

8 kHz, PCM, Windows WAV format

8 kHz, mulaw (type 1), AU header

8 kHz, PCM (type 3), AU header

Absolute hertz: 50 to 300

Relative hertz: -50 to +100

Relative percentages: -30% to +30%

Relative semitones: +5st to -5st

position is the percentage of the period of the

target is the target pitch at that position. See

Absolute hertz: 0 to 100

Fixed percentages: 75% to 150%

Descriptive: loud, medium, soft, silent

Fixed percentages: 0% to 100%

For WAV files:

8 kHz, mulaw, unheadered

8 kHz, mulaw, Sphere header

8 kHz, PCM, Windows WAV format

8 kHz, mulaw (type 1), AU header

8 kHz, PCM (type 3), AU header

You can use either / or \ for path components

The only authority Vocalizer handles for file URIs is localhost or