Method of adding vocabulary to a speech recognition system

• US 6,334,102 B1
• Filed: 09/13/1999
• Issued: 12/25/2001
• Est. Priority Date: 09/13/1999
• Status: Expired due to Term

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Lewis, James R., Ortega, Kerry A.
• Primary Examiner(s): Dorvil, Richemond
• Application Number:
  US09/394,730
• Time in Patent Office: 834 Days
• Field of Search: 704/243, 704/251, 704/256, 704/255, 704/257, 704/252, 704/247, 704/244, 704/270, 704/272, 704/277, 704/278
• US Class Current: 704/255
• CPC Class Codes: G10L 15/197 Probabilistic grammars, e.g. word n-grams

Abstract

A method and a system for use in a computer speech recognition system for adding new vocabulary by using language model statistics corresponding to an existing vocabulary word. The method involves a series of steps including receiving a user input identifying a first word for which no language model statistics exist in the speech recognition system. The first word is for inclusion within the existing vocabulary of the speech recognition system. In response to a second user input identifying a second word for which language model statistics exist in the speech recognition system, recalling from a computer memory the language model statistics for the second word. The speech recognition system then automatically creates language model statistics for the first word by duplicating the language model statistics of the second word and replacing each occurrence of the second word in the duplicated language model statistics with the first word.

Claims

1. In a computer speech recognition system, a method for adding new vocabulary by using language model statistics corresponding to an existing vocabulary word, comprising the steps of:

receiving a first user input identifying a first word for which no language model statistics exist in said speech recognition system, said first word for inclusion within the existing vocabulary of said speech recognition system;

in response to a second user input identifying a second word for which language model statistics exist in said speech recognition system, recalling from a computer memory said language model statistics for said second word; and

,

automatically creating language model statistics for said first word by duplicating said language model statistics of said second word and replacing each occurrence of said second word in said duplicated language model statistics with said first word.

2. The method according to claim 1, further comprising the step of:

receiving a user input specifying a relative frequency of said first word in relation to said second word.

3. The method according to claim 2, further comprising the step of:

automatically updating said language model statistics for said first word by modifying frequency values in said language model statistics for said first word according to said user specified relative frequency of said first word.

4. The method according to claim 3 further comprising the steps of:

presenting said user with at least one of a bigram and trigram from said language model statistics for said first word in a user readable format; and

,

receiving user input specifying modifications to said bigrams and trigrams from said language model statistics for said first word, for inclusion in a language model of said speech recognition system.

5. The method according to claim 4, wherein each of said user inputs is a spoken utterance.

6. The method according to claim 3, wherein said updating step comprises updating a frequency value of each unigram, bigram, and trigram containing said first word according to said user specified relative frequency of said first word.

7. The method according to claim 1, wherein each of said user inputs is a spoken utterance.

8. The method according to claim 1, wherein said first word and said second word are related in meaning.

9. The method according to claim 1 wherein said language model statistics for said second word are comprised of each unigram, bigram, and trigram containing said second word and a frequency value for each said unigram, bigram, and trigram.

10. The method according to claim 1, wherein said creating step comprises:

duplicating each unigram, bigram, and trigram containing said second word and a frequency value for each said unigram, bigram, and trigram; and

,

replacing each occurrence of said second word in each said unigram, bigram, and trigram containing said second word with said first word.

11. In a computer speech recognition system, a system for adding new vocabulary by using language model statistics corresponding to an existing vocabulary word, comprising:

means for receiving a first user input identifying a first word for which no language model statistics exist in said speech recognition system, said first word for inclusion within the existing vocabulary of said speech recognition system;

means for recalling from a computer memory, language model statistics for a second word in response to a second user input identifying said second word for which said language model statistics exist in said speech recognition system; and

,

means for automatically creating language model statistics for said first word by duplicating said language model statistics of said second word and replacing each occurrence of said second word in said duplicated language model statistics with said first word.

12. The system according to claim 11, further comprising:

means for receiving a user input specifying a relative frequency of said first word in relation to said second word.

13. The system according to claim 12, further comprising:

means for automatically updating said language model statistics for said first word by modifying frequency values in said language model statistics for said first word according to said user specified relative frequency of said first word.

14. The system according to claim 13, further comprising:

means for presenting said user with at least one of a bigram and trigram from said language model statistics for said first word in a user readable format; and

,

means for receiving user input specifying modifications to said bigrams and trigrams from said language model statistics for said first word, for inclusion in a language model of said speech recognition system.

15. The system according to claim 14, wherein each of said user inputs is a spoken utterance.

16. The system according to claim 13, wherein said updating means further comprises means for updating a frequency value of each unigram, bigram, and trigram containing said first word according to said user specified relative frequency of said first word.

17. The system according to claim 11, wherein each of said user inputs is a spoken utterance.

18. The system according to claim 11, wherein said first word and said second word are related in meaning.

19. The system according to claim 11, wherein said language model statistics for said second word are comprised of each unigram, bigram, and trigram containing said second word and a frequency value for each said unigram, bigram, and trigram.

20. The system according to claim 11, wherein said means for automatically creating language model statistics further comprises:

means for duplicating each unigram, bigram, and trigram containing said second word and a frequency value for each said unigram, bigram, and trigram; and

,

means for replacing each occurrence of said second word in each said unigram, bigram, and trigram containing said second word with said first word.

21. A machine readable stored thereon a computer program having a plurality of code sections executable by a machine for causing the machine to perform the steps of:

receiving a first user input identifying a first word for which no language model statistics exist in said speech recognition system, said first word for inclusion within the existing vocabulary of said speech recognition system;

in response to a second user input identifying a second word for which language model statistics exist in said speech recognition system, recalling from a computer memory said language model statistics for said second word; and

,

automatically creating language model statistics for said first word by duplicating said language model statistics of said second word and replacing each occurrence of said second word in said duplicated language model statistics with said first word.