Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Method for Differential Text-to-Speech Acronym Pronunciation

IP.com Disclosure Number: IPCOM000111248D
Original Publication Date: 1994-Feb-01
Included in the Prior Art Database: 2005-Mar-26
Document File: 2 page(s) / 38K

Publishing Venue

IBM

Related People

Salahshour, A: AUTHOR [+2]

Abstract

Current Text-to-Speech applications, typically, attempt to pronounce words which are acronyms. Such pronunciations are confusing to listeners. More recent Text-to-Speech systems, pronounce each letter of the acronym if the acronym is capitalized. As many users know some acronyms are best enunciated letter by letter, e.g., IBM, while others are best enunciated as a single word. In addition, some services allow the user to manually stipulate how certain words should be pronounced, however this alternative would require the user know in advance what acronyms are going to be encountered within the specified text. It is desirable to provide a mechanism by which a Text-to-Speech service can use differing enunciation techniques depending on the spelling of the acronym.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 86% of the total text.

Method for Differential Text-to-Speech Acronym Pronunciation

      Current Text-to-Speech applications, typically, attempt to
pronounce words which are acronyms.  Such pronunciations are
confusing to listeners.  More recent Text-to-Speech systems,
pronounce each letter of the acronym if the acronym is capitalized.
As many users know some acronyms are best enunciated letter by
letter, e.g., IBM, while others are best enunciated as a single word.
In addition, some services allow the user to manually stipulate how
certain words should be pronounced, however this alternative would
require the user know in advance what acronyms are going to be
encountered within the specified text.  It is desirable to provide a
mechanism by which a Text-to-Speech service can use differing
enunciation techniques depending on the spelling of the acronym.

      Disclosed is a method by which vowels and consonants are
compared to a soundex based table.  Each letter is associated to a
particular numeric value, similar to soundex searching routines.
Single consonants, followed by vowels are used with other soundex
patterns to create an enunciation table.  Upon the Text-to-Speech
service locating an acronym, e.g., a word with all capital letters, a
service interrogates the soundex representation of the acronym, if
the acronym meets any of the specified criteria within the
enunciation table the acronym is pronounced accordingly, otherwise
each letter is enunciated.  It should be noted tha...