End-of-utterance determination for voice processing

• US 6,496,799 B1
• Filed: 06/13/2000
• Issued: 12/17/2002
• Est. Priority Date: 12/22/1999
• Status: Expired due to Term

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Pickering, John Brian
• Primary Examiner(s): Dorvil, Richemond
• Assistant Examiner(s): Nolan, Daniel A.
• Application Number:
  US09/592,798
• Time in Patent Office: 917 Days
• Field of Search: 704/235, 704/251, 704/257, 704/249, 704/250, 704/260, 704/253
• US Class Current: 704/235
• CPC Class Codes: G10L 15/04 Segmentation; Word boundary detection
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning

Abstract

A voice processing system receives spoken input from a user, which is subjected to speech recognition to convert the spoken input into a text equivalent. The semantic and/or prosodic properties of the spoken input are then analysed, and based on the results of this analysis, it is determined whether or not the user input has effectively completed. Thus for example if a semantic analysis indicates that the user is no longer saying anything useful, then the system may decide to interrupt the user and play a further prompt to him or her, clarifying what extra input is required.

Claims

1. A method of operating a voice processing system comprising the steps of:

receiving spoken input from a user;

performing speech recognition to convert said spoken input into text equivalent;

analysing at least one semantic or prosodic property of said spoken input by looking for task words in the text equivalent of the spoken input; and

responsive to said analysis, determining that the user input has effectively completed if there has not been a task word for more than a predetermined period of time.

2. The method of claim 1, wherein responsive to a determination that the user input has effectively completed, the system plays a prompt to the user.

3. The method of claim 2, wherein said prompt is played to the user when there is some break in the spoken input.

4. The method of claim 3, wherein said break is insufficient to trigger a general time-out for user input into the voice processing system.

5. The method of claim 1, wherein said step of analysing at least one semantic or prosodic property of said spoken input comprises looking for a prolonged pitch fall indicating effective completion of the spoken input.

6. The method of claim 1, wherein said step of analysing at least one semantic or prosodic property of said spoken input comprises looking for a reset of the pitch excursion envelope indicating effective completion of the spoken input.

7. The method of claim 1, wherein said step of analysing at least one semantic or prosodic property of said spoken input comprises looking for a question in the spoken input.

8. The method of claim 1, in which said voice processing system and said user communicate with each other over a telephone network, whereby said spoken input is received over a telephone connection.

9. A voice processing apparatus comprising:

an input device to receive spoken input from a user;

a speech recognition unit to convert said spoken input into text equivalent; and

means for analysing at least one semantic or prosodic property of said spoken input by for looking for task words in the text equivalent of the spoken input; and

wherein responsive to said analysis, it is determined that the user input has effectively completed if there has not been a task word for more than a predetermined period of time.

10. The apparatus of claim 9, wherein responsive to a determination that the user input has effectively completed, the apparatus plays a prompt to the user.

11. The apparatus of claim 10, wherein said prompt is played to the user when there is some break in the spoken input.

12. The apparatus of claim 11, wherein said break is insufficient to trigger a general time-out for user input into the voice processing apparatus.

13. The apparatus of claim 9, wherein said means for analysing at least one semantic or prosodic property of said spoken input comprises means for looking for a prolonged pitch fall indicating effective completion of the spoken input.

14. The apparatus of claim 9, wherein said means for analysing at least one semantic or prosodic property of said spoken input comprises means for looking for a reset of the pitch excursion envelope indicating effective completion of the spoken input.

15. The apparatus of claim 9, wherein said means for analysing at least one semantic or prosodic property of said spoken input comprises means for looking for a question in the spoken input.

16. The apparatus of claim 9, in which said voice processing apparatus and said user communicate with each other over a telephone network, whereby said spoken input is received over a telephone connection.

17. A computer readable medium containing instructions readable by a computer system operating as a voice processing system, said instructions including instructions for causing the computer system to perform the following steps:

receiving spoken input from a user;

performing speech recognition to convert said spoken input into text equivalent;

analysing at least one semantic or prosodic property of said spoken input by looking for task words in the text equivalent of the spoken input; and

responsive to said analysis, determining that the user input has effectively completed if there has not been a task word for more than a predetermined period of time.

18. The medium of claim 17, wherein responsive to a determination that the user input has effectively completed, the system plays a prompt to the user.

19. The medium of claim 18, wherein said prompt is played to the user when there is some break in the spoken input.

20. The medium of claim 19, wherein said break is insufficient to trigger a general time-out for user input into the voice processing system.

21. The medium of claim 17, wherein said analysing at least one semantic or prosodic property of said spoken input comprises looking for a prolonged pitch fall indicating effective completion of the spoken input.

22. The medium of claim 17, wherein said analysing at least one semantic or prosodic property of said spoken input comprises looking for a reset of the pitch excursion envelope indicating effective completion of the spoken input.

23. The medium of claim 17, wherein said analysing at least one semantic or prosodic property of said spoken input comprises looking for a question in the spoken input.

24. The medium of claim 17, in which said voice processing system and said user communicate with each other over a telephone network, whereby said spoken input is received over a telephone connection.