Method of transcribing text from computer voice mail

• US 6,775,651 B1
• Filed: 05/26/2000
• Issued: 08/10/2004
• Est. Priority Date: 05/26/2000
• 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): CHAWAN, VIJAY B
• Application Number:
  US09/579,985
• Time in Patent Office: 1,537 Days
• Field of Search: 704/235, 704/260, 704/270, 704/201, 704/231, 704/211, 704/246, 704/270.1, 704/278, 704/275, 379/88.01, 379/88.02, 379/88.04, 379/88.22, 379/88.13, 379/88.14, 379/93.24, 379/88.23, 379/100.08, 379/67.1, 379/80, 379/80.24, 379/93.03
• US Class Current: 704/235
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: H04M 2201/40 using speech recognition
• CPC Class Codes: H04M 2201/60 Medium conversion
• CPC Class Codes: H04M 2242/22 Automatic class or number identification arrangements
• CPC Class Codes: H04M 3/42042 Notifying the called party of information on the calling party details within substation equipment H04M1/57, signalling details H04Q3/72
• CPC Class Codes: H04M 3/5307 for recording messages comprising any combination of audio and non-audio components

Abstract

The invention concerns a method and a system for transcribing a voice mail message. The method of the invention involves a plurality of steps including, first providing a computer voice mail message stored in an audio file to a computer speech recognition system and, second, submitting the computer voice mail message to a speaker identification process in the speech recognition system. Notably, the speaker identification process can identify an enrolled speaker as a source of the computer voice mail message. Finally, responsive to the identification of the enrolled speaker, the computer voice mail message can be submitted to a speech conversion process in the speech recognition system. The speech conversion process can perform speech-to-text conversion of the computer voice mail message using speaker enrollment data corresponding to the identified enrolled speaker. Furthermore, the speech-to-text conversion can produce a transcription of the computer voice mail message. In one embodiment of the present invention, the transcription further can be displayed.

Claims

1. A method of transcribing a computer voice mail comprising:

providing a computer voice mail message stored in an audio file to a computer speech recognition system;

submitting said computer voice mail message to a speaker identification process in said speech recognition system, said speaker identification process identifying an enrolled speaker as a source of said computer voice mail message; and

,

responsive to said identification of said enrolled speaker, submitting said computer voice mail message to a speech conversion process in said speech recognition system, said speech conversion process performing speaker-dependent speech-to-text conversion of said computer voice mail message using speaker enrollment data corresponding to said identified enrolled speaker;

said speech-to-text conversion producing a transcription of said computer voice mail message.

2. The method of claim 1, further comprising the steps of:

displaying said transcription.

3. The method of claim 1, wherein said speaker identification process comprises the steps of:

identifying an enrolled speaker having speaker enrollment data as a source of said voice mail message using text-independent speaker identification.

4. The method of claim 1, wherein said speaker identification process further comprises the steps of:

if said speaker identification process fails to identify an enrolled speaker as a source of said computer voice mail message, creating a speaker enrollment;

associating said created speaker enrollment with a non-enrolled speaker; and

,

identifying said associated speaker as a source of said voice mail message.

5. The method of claim 4, wherein said step of creating an enrollment comprises the step of:

performing an unsupervised enrollment of said associated speaker.

6. The method of claim 1, wherein said speaker identification process comprises the steps of:

providing to a user a list of enrolled speakers, each enrolled speaker having corresponding enrollment data;

accepting a selection by said user of one of said enrolled speakers in said list; and

,

identifying said selected enrolled speaker as a source of said voice mail message.

7. A system of transcribing a voice mail message comprising:

a voice mail system for recording a voice mail message spoken by a caller;

a speaker identification processor for identifying a source speaker associated with said recorded voice mail message; and

,

a speech recognition system for performing speaker-dependent speech-to-text conversion of said recorded voice mail message using speaker enrollment data corresponding to said identified source speaker associated with said recorded voice mail message, said speech-to-text conversion producing a transcription of said voice mail message.

8. The system of claim 7, further comprising:

display means for displaying said transcription.

9. The system of claim 8, wherein said display means is selected from the group of a printer for printing said transcription and a user interface for visually displaying said transcription.

10. The system of claim 7, wherein said speaker identification processor implements a text-independent speaker identification technique.

11. The system of claim 7, further comprising:

an unsupervised enrollment processor for creating speaker enrollment data associated with a source of said voice mail message not identified by said speaker identification processor;

said speech recognition system performing said speech-to-text conversion of a voice mail message spoken by said unknown speaker using said created speaker enrollment data.

12. A machine readable storage, having stored thereon a computer program for transcribing a voice mail message, said computer program having a plurality of code sections executable by a machine for causing the machine to perform the steps of:

providing a computer voice mail message stored in an audio file to a computer speech recognition system;

submitting said computer voice mail message to a speaker identification process in said speech recognition system, said speaker identification process identifying an enrolled speaker as a source of said computer voice mail message; and

,

responsive to said identification of said enrolled speaker, submitting said computer voice mail message to a speech conversion process in said speech recognition system, said speech conversion process performing speaker-dependent speech-to-text conversion of said computer voice mail message using speaker enrollment data corresponding to said identified enrolled speaker;

said speech-to-text conversion producing a transcription of said computer voice mail message.

13. The machine readable storage of claim 12, said computer program further comprising a plurality of code sections executable by a machine for further causing the machine to perform the step of:

displaying said transcription.

14. The machine readable storage of claim 12, wherein said speaker identification process comprises the step of:

identifying an enrolled speaker having speaker enrollment data as a source of said voice mail message using text-independent speaker identification.

15. The machine readable storage of claim 12, wherein said speaker identification process further comprises the steps of:

if said speaker identification process fails to identify an enrolled speaker as a source of said computer voice mail message, creating a speaker enrollment;

associating said created speaker enrollment with a non-enrolled speaker; and

,

identifying said associated speaker as a source of said voice mail message.

16. The machine readable storage of claim 15, wherein said step of creating an enrollment comprises the step of:

performing an unsupervised enrollment of said associated speaker.

17. The machine readable storage of claim 12, wherein said speaker identification process comprises the steps of:

providing to a user a list of enrolled speakers, each enrolled speaker having corresponding enrollment data;

accepting a selection by said user of one of said enrolled speakers in said list; and

,

identifying said selected enrolled speaker as a source of said voice mail message.

18. A method of transcribing a computer voice mail comprising:

providing a computer voice mail message stored in an audio file to a computer speech recognition system;

utilizing prosodic information within said computer voice mail message to automatically identify an enrolled speaker as a source of said computer voice mail message; and

,

responsive to said identification of said enrolled speaker, submitting said computer voice mail message to a speech conversion process in said speech recognition system, said speech conversion process performing speech-to-text conversion of said computer voice mail message using speaker enrollment data corresponding to said identified enrolled speaker;

said speech-to-text conversion producing a transcription of said computer voice mail message.

19. The method of claim 18, said speech conversion process further comprising the step of performing speaker-dependent speech conversion to produce said transcription.

20. A machine readable storage, having stored thereon a computer program for transcribing a voice mail message, said computer program having a plurality of code sections executable by a machine for causing the machine to perform the steps of:

providing a computer voice mail message stored in an audio file to a computer speech recognition system;

utilizing prosodic information within said computer voice mail message to automatically identify an enrolled speaker as a source of said computer voice mail message; and

,

responsive to said identification of said enrolled speaker, submitting said computer voice mail message to a speech conversion process in said speech recognition system, said speech conversion process performing speech-to-text conversion of said computer voice mail message using speaker enrollment data corresponding to said identified enrolled speaker;

said speech-to-text conversion producing a transcription of said computer voice mail message.

21. The machine readable storage of claim 20, said speech conversion process further comprising the step of performing speaker-dependent speech conversion to produce said transcription.