Real time automatic caller speech profiling

• US 8,824,641 B2
• Filed: 10/18/2013
• Issued: 09/02/2014
• Est. Priority Date: 11/10/2009
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Daye, Sheri G., Jaiswal, Peeyush, Wang, Fang
• Primary Examiner(s): ISLAM, MOHAMMAD K
• Application Number:
  US14/057,600
• Publication Number:
  US 20140064464A1
• Time in Patent Office: 319 Days
• Field of Search: 704/9, 704/246, 704/257, 704/273, 704/250, 709/204
• US Class Current: 379/88.01
• CPC Class Codes: G10L 15/06 Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice G10L15/14 takes precedence
• CPC Class Codes: G10L 2015/0631 Creating reference templates; Clustering
• CPC Class Codes: H04M 2201/40 using speech recognition speech recognition per se G10L15/00
• CPC Class Codes: H04M 2203/2061 Language aspects
• CPC Class Codes: H04M 3/4936 Speech interaction details speech recognition per se G10L15/00

Abstract

A computer-implemented method, system, and/or computer program product recognizes a voice input received by an interactive voice recognition system. An unintelligible voice input, which is unintelligible due to background noise at the caller'"'"'s location, is received from a caller. Upon determining that the unintelligible voice input does not match any entry from a lexicon of known voice inputs, the unintelligible voice input is stored in a no-match-based user profiles database, and the caller is prompted to transmit a non-verbal electronic input to clarify the unintelligible voice input. This non-verbal electronic input is used to match the unintelligible voice input with a specific known voice input from the lexicon of known voice inputs, which is updated with the unintelligible voice input. The updated lexicon is subsequently used in future phone calls from the user to match the unintelligible voice input to the user-requested activity for the caller.

Claims

1. A method of recognizing a voice input received by an interactive voice recognition system, the method comprising:

receiving an unintelligible voice input from a caller;

digitizing, by one or more processors, the unintelligible voice input to generate an unintelligible digitized voice input;

determining, by one or more processors, that the unintelligible digitized voice input is unintelligible to a computer due to background noise at a caller environment of the caller;

in response to determining that the unintelligible digitized voice input does not match any entry from a lexicon of known digitized voice inputs, storing the unintelligible digitized voice input in a no-match-based user profiles database;

prompting the caller to transmit a secondary input to clarify the unintelligible digitized voice input, wherein the secondary input is a non-verbal electronic input that identifies a user-requested activity that the caller was requesting when transmitting the voice input;

utilizing the non-verbal electronic input to match the unintelligible digitized voice input with a specific known digitized voice input from the lexicon of known digitized voice inputs;

comparing the unintelligible digitized voice input to the specific known digitized voice input to derive a voice input match percentage, wherein the voice input match percentage describes how closely the unintelligible digitized voice input and the specific known digitized voice input match one another;

in response to the voice input match percentage exceeding a predetermined value, updating the lexicon of known digitized voice inputs with the unintelligible digitized voice input, wherein the unintelligible digitized voice input is matched to the specific known digitized voice input from the lexicon of known digitized voice inputs;

matching the unintelligible digitized voice input to the user-requested activity in a database that is specific for the caller;

receiving the unintelligible voice input in a subsequent phone call from the user and converting the unintelligible voice input in the subsequent phone call into the unintelligible digitized voice input; and

matching the unintelligible digitized voice input from the subsequent phone call to the user-requested activity for the caller.

2. The method of claim 1, wherein the secondary input further comprises another caller voice input from the caller.

3. The method of claim 2, further comprising:

in response to said another caller voice input from the caller matching the specific known digitized voice input, discarding said another caller voice input from the lexicon of known digitized voice inputs.

4. The method of claim 1, wherein the secondary input is a caller'"'"'s response to a menu of input options, wherein one of the input options identifies an input whose definition matches that of the specific known digitized voice input from the lexicon of known digitized voice inputs.

5. The method of claim 1, wherein the lexicon of known digitized voice inputs is exclusive to the caller.

6. The method of claim 1, wherein the lexicon of known digitized voice inputs is non-exclusive to the caller.

7. The method of claim 1, further comprising:

matching the unintelligible voice input to an interactive voice recognition activity; and

utilizing a match between the unintelligible voice input and the interactive voice recognition activity to match the unintelligible digitized voice input to the specific known digitized voice input from the lexicon of known digitized voice inputs.

8. The method of claim 1, wherein updating the lexicon of known digitized voice inputs with the unintelligible digitized voice input creates an updated lexicon of known digitized voice inputs, and wherein the method further comprises:

receiving a new call from the caller, wherein the caller transmits a new caller voice input;

converting the new caller voice input into a new caller digitized voice input; and

matching the new caller digitized voice input with a known digitized voice input from the updated lexicon of known digitized voice inputs.

9. A computer program product for recognizing a voice input received by an interactive voice recognition system, the computer program product comprising a computer readable storage medium having program code embodied therewith, the program code readable and executable by a processor to:

receive an unintelligible voice input from a caller;

digitize the unintelligible voice input to generate an unintelligible digitized voice input;

determine that the unintelligible digitized voice input is unintelligible to a computer due to background noise at a caller environment of the caller;

in response to determining that the unintelligible digitized voice input does not match any entry from a lexicon of known digitized voice inputs, store the unintelligible digitized voice input in a no-match-based user profiles database;

prompt the caller to transmit a secondary input to clarify the unintelligible digitized voice input, wherein the secondary input is a non-verbal electronic input that identifies a user-requested activity that the caller was requesting when transmitting the voice input;

utilize the non-verbal electronic input to match the unintelligible digitized voice input with a specific known digitized voice input from the lexicon of known digitized voice inputs;

compare the unintelligible digitized voice input to the specific known digitized voice input to derive a voice input match percentage, wherein the voice input match percentage describes how closely the unintelligible digitized voice input and the specific known digitized voice input match one another;

in response to the voice input match percentage exceeding a predetermined value, update the lexicon of known digitized voice inputs with the unintelligible digitized voice input, wherein the unintelligible digitized voice input is matched to the specific known digitized voice input from the lexicon of known digitized voice inputs;

match the unintelligible digitized voice input to the user-requested activity in a database that is specific for the caller;

receive the unintelligible voice input in a subsequent phone call from the user and converting the unintelligible voice input in the subsequent phone call into the unintelligible digitized voice input; and

match the unintelligible digitized voice input from the subsequent phone call to the user-requested activity for the caller.

10. The computer program product of claim 9, wherein the secondary input further comprises another caller voice input from the caller.

11. The computer program product of claim 10, wherein the program code is further readable and executable by a processor to:

in response to said another caller voice input from the caller matching the specific known digitized voice input, discard said another caller voice input from the lexicon of known digitized voice inputs.

12. The computer program product of claim 9, wherein the secondary input is a caller'"'"'s response to a menu of input options, wherein one of the input options identifies an input whose definition matches that of the specific known digitized voice input from the lexicon of known digitized voice inputs.

13. The computer program product of claim 9, wherein the lexicon of known digitized voice inputs is exclusive to the caller.

14. The computer program product of claim 9, wherein the lexicon of known digitized voice inputs is non-exclusive to the caller.

15. The computer program product of claim 9, wherein the program code is further readable and executable by a processor to:

match the unintelligible voice input to an interactive voice recognition activity; and

utilize a match between the unintelligible voice input and the interactive voice recognition activity to match the unintelligible digitized voice input to the specific known digitized voice input from the lexicon of known digitized voice inputs.

16. The computer program product of claim 9, wherein updating the lexicon of known digitized voice inputs with the unintelligible digitized voice input creates an updated lexicon of known digitized voice inputs, and wherein the program code is further readable and executable by a processor to:

receive a new call from the caller, wherein the caller transmits a new caller voice input;

convert the new caller voice input into a new caller digitized voice input; and

match the new caller digitized voice input with a known digitized voice input from the updated lexicon of known digitized voice inputs.

17. A computer system comprising:

a processor, a computer readable memory, and a non-transitory computer readable storage media;

first program instructions to receive an unintelligible voice input from a caller;

second program instructions to digitize the unintelligible voice input to generate an unintelligible digitized voice input;

third program instructions to determine that the unintelligible digitized voice input is unintelligible to a computer due to background noise at a caller environment of the caller;

fourth program instructions to, in response to determining that the unintelligible digitized voice input does not match any entry from a lexicon of known digitized voice inputs, store the unintelligible digitized voice input in a no-match-based user profiles database;

fifth program instructions to prompt the caller to transmit a secondary input to clarify the unintelligible digitized voice input, wherein the secondary input is a non-verbal electronic input that identifies a user-requested activity that the caller was requesting when transmitting the voice input;

sixth program instructions to utilize the non-verbal electronic input to match the unintelligible digitized voice input with a specific known digitized voice input from the lexicon of known digitized voice inputs;

seventh program instructions to compare the unintelligible digitized voice input to the specific known digitized voice input to derive a voice input match percentage, wherein the voice input match percentage describes how closely the unintelligible digitized voice input and the specific known digitized voice input match one another;

eighth program instructions to, in response to the voice input match percentage exceeding a predetermined value, update the lexicon of known digitized voice inputs with the unintelligible digitized voice input, wherein the unintelligible digitized voice input is matched to the specific known digitized voice input from the lexicon of known digitized voice inputs;

ninth program instructions to match the unintelligible digitized voice input to the user-requested activity in a database that is specific for the caller;

tenth program instructions to receive the unintelligible voice input in a subsequent phone call from the user and converting the unintelligible voice input in the subsequent phone call into the unintelligible digitized voice input; and

eleventh program instructions to match the unintelligible digitized voice input from the subsequent phone call to the user-requested activity for the caller; and

wherein the first, second, third, fourth, fifth, sixth, seventh, eighth, ninth, tenth, and eleventh program instructions are stored on the non-transitory computer readable storage media for execution by the processor via the computer readable memory.

18. The computer system of claim 17, wherein the secondary input further comprises another caller voice input from the caller.

19. The computer system of claim 17, wherein the secondary input is a caller'"'"'s response to a menu of input options, wherein one of the input options identifies an input whose definition matches that of the specific known digitized voice input from the lexicon of known digitized voice inputs.

20. The computer system of claim 17, wherein the lexicon of known digitized voice inputs is non-exclusive to the caller.