Pronunciation Discovery for Spoken Words

Abstract

A method for a portable device includes receiving a spoken utterance of a word or phrase, generating a plurality of alternative pronunciations of the spoken utterance, scoring one or more pronunciations of the plurality of alternative pronunciations using the spoken utterance, and updating a lexicon with at least one scored pronunciation.

What is claimed is: 
     
         1 . A method comprising:
 receiving a spoken utterance of a word or phrase;   generating a plurality of alternative pronunciations of the spoken utterance;   scoring one or more pronunciations of the plurality of alternative pronunciations using the spoken utterance; and   updating a lexicon with at least one scored pronunciation.   
     
     
         2 . The method of  claim 1 , wherein the at least one scored pronunciation is the highest scoring pronunciation of the scored pronunciations. 
     
     
         3 . The method of  claim 1 , further comprising using a finite-state machine to generate the plurality of alternative pronunciations. 
     
     
         4 . The method of  claim 1 , wherein updating the lexicon includes replacing an existing pronunciation with the at least one scored pronunciation. 
     
     
         5 . The method of  claim 1 , wherein updating the lexicon includes adding a phonetic representation of the at least one scored pronunciation to an existing representation. 
     
     
         6 . The method of  claim 1 , wherein the alternative pronunciations are generated by searching a neighborhood of pronunciations about an initial pronunciation of the spoken utterance. 
     
     
         7 . A system comprising:
 at least one processor; and   a memory device operatively connected to the at least one processor;   wherein, responsive to execution of program instructions accessible to the at least one processor, the at least one processor is configured to:
 receive a spoken utterance of a word or phrase; 
 generate a plurality of alternative pronunciations of the spoken utterance; 
 score one or more pronunciations of the plurality of alternative pronunciations using the spoken utterance; and 
 update a lexicon with at least one scored pronunciation. 
   
     
     
         8 . The system of  claim 7 , wherein the at least one scored pronunciation is the highest scoring pronunciation of the scored pronunciations. 
     
     
         9 . The system of  claim 7 , wherein the at least one processor is configured to use a finite-state machine to generate the plurality of alternative pronunciations. 
     
     
         10 . The system of  claim 7 , wherein the at least one processor is configured to update the lexicon by replacing an existing pronunciation with the at least one scored pronunciation. 
     
     
         11 . The system of  claim 7 , wherein the at least one processor is configured to update the lexicon by adding a phonetic representation of the at least one scored pronunciation to an existing representation. 
     
     
         12 . The system of  claim 7 , wherein the at least one processor is configured to generate the alternative pronunciations by searching a neighborhood of pronunciations about an initial pronunciation of the spoken utterance. 
     
     
         13 . A computer program product encoded in a non-transitory computer-readable medium, which when executed by a computer causes the computer to perform the following operations:
 receiving a spoken utterance of a word or phrase;   generating a plurality of alternative pronunciations of the spoken utterance;   scoring one or more pronunciations of the plurality of alternative pronunciations using the spoken utterance; and   updating a lexicon with at least one scored pronunciation.   
     
     
         14 . The computer program product of  claim 13 , wherein the at least one scored pronunciation is the highest scoring pronunciation of the scored pronunciations. 
     
     
         15 . The computer program product of  claim 13 , wherein the computer uses a finite-state machine to generate the plurality of alternative pronunciations. 
     
     
         16 . The computer program product of  claim 13 , wherein the computer updates the lexicon by replacing an existing pronunciation with the at least one scored pronunciation. 
     
     
         17 . The computer program product of  claim 13 , wherein the computer updates the lexicon by adding a phonetic representation of the at least one scored pronunciation to an existing representation. 
     
     
         18 . The computer program product of  claim 13 , wherein the computer generates the alternative pronunciations by searching a neighborhood of pronunciations about an initial pronunciation of the spoken utterance.