Granted Patent B2

US 12,334,079 · App. 17/447,579 · Granted Jun 17, 2025

Inventors: Aaron K. Baughman (Cary, NC); Shikhar Kwatra (San Jose, CA); Saurabh Yadav (Bangalore, IN); Eric Jeffery (Monument, CO)

Assignee: INTERNATIONAL BUSINESS MACHINES CORPORATION

G10L17/06G06F40/166G06F40/30G10L25/18G10L25/24H04M3/436

Abstract

Claims (50)

1. A processor-implemented method for identifying a spam call, the method comprising:

capturing an audio data associated with a voice of a user;

determining a user frequency spectrum and a user voice pattern associated with the voice of the user;

identifying, using natural language processing (NLP), one or more first topics communicated by the user in a previous conversation including the user as a caller;

based on detecting a call from the caller to a receiver, identifying the caller as the user based on matching the user frequency spectrum to a caller frequency spectrum determined from a current conversation in the call and the user voice pattern to a caller voice pattern determined from the current conversation in the call;

responsive to identifying the caller as the user, displaying on a computing device of the receiver, a stored trust score associated with the caller, wherein the stored trust score indicates to the receiver whether the caller is trustworthy based on a previous call between the caller and a different receiver, wherein the previous call includes the previous conversation; and

determining an updated trust score for the caller relative to the stored trust score based on comparing the one or more first topics communicated by the caller in the previous conversation to one or more second topics communicated by the caller in the current conversation with the receiver to identify differences between the one or more first topics and the one or more second topics communicated by a same person, wherein the updated trust score is determined based on a plurality of flag counters, wherein at least one of the plurality of flag counters are generated based on a discrepancy by the caller in a user name pronunciation as compared to a stored user name pronunciation.

2. The method of claim 1 , wherein the user frequency spectrum and the caller frequency spectrum are determined by a non-Bayesian Gaussian mixture model.

3. The method of claim 1 , wherein the user voice pattern and the caller voice pattern are determined by a Mel Frequency Cepstral Coefficients algorithm.

4. The method of claim 1 , wherein identifying, using NLP, the one or more first topics associated with the user from the previous conversation including the user further comprises:

determining the one or more first topics by an automatic text summarizer and a Latent Dirichlet Allocation model.

5. The method of claim 1 , further comprising:

storing the user frequency spectrum, the user voice pattern, the one or more first topics, the stored trust score, and the updated trust score under a user ID associated with the user.

6. The method of claim 1 , wherein the stored trust score and the updated trust score are further determined by a Dirichlet distribution model having parameters obtained by training the Dirichlet distribution model with training samples of callers.

7. The method of claim 1 , further comprising:

requesting the user to opt in for a service;

based on determining the user refused to opt in for the service, setting the stored trust score to a default value; and

based on determining the stored trust score is set to the default value, causing the computing device of the receiver of the call to display that the user refused to opt in for the service, wherein the user is the caller.

8. A computer system for identifying a spam call, the computer system comprising: one or more processors, one or more computer-readable memories, one or more computer readable tangible storage medium, and program instructions stored on at least one of the one or more tangible storage medium for execution by at least one of the one or more processors via at least one of the one or more memories, wherein the computer system is capable of performing a method comprising:

capturing an audio data associated with a voice of a user;

determining a user frequency spectrum and a user voice pattern associated with the voice of the user;

identifying, using natural language processing (NLP), one or more first topics communicated by the user in a previous conversation including the user as a caller;

based on detecting a call from the caller to a receiver, identifying the caller as the user based on matching the user frequency spectrum to a caller frequency spectrum determined from a current conversation in the call and the user voice pattern to a caller voice pattern determined from the current conversation in the call;

responsive to identifying the caller as the user, displaying on a computing device of the receiver, a stored trust score associated with the caller, wherein the stored trust score indicates to the receiver whether the caller is trustworthy based on a previous call between the caller and a different receiver, wherein the previous call includes the previous conversation; and

determining an updated trust score for the caller relative to the stored trust score based on comparing the one or more first topics communicated by the caller in the previous conversation to one or more second topics communicated by the caller in the current conversation with the receiver to identify differences between the one or more first topics and the one or more second topics communicated by a same person, wherein the updated trust score is determined based on a plurality of flag counters, wherein at least one of the plurality of flag counters are generated based on a discrepancy by the caller in a user name pronunciation as compared to a stored user name pronunciation.

9. The computer system of claim 8 , wherein the user frequency spectrum and the caller frequency spectrum are determined by a non-Bayesian Gaussian mixture model.

10. The computer system of claim 8 , wherein the user voice pattern and the caller voice pattern are determined by a Mel Frequency Cepstral Coefficients algorithm.

11. The computer system of claim 8 , wherein identifying, using NLP, the one or more first topics associated with the user from the previous conversation including the user further comprises:

determining the one or more first topics by an automatic text summarizer and a Latent Dirichlet Allocation model.

12. The computer system of claim 8 , further comprising:

storing the user frequency spectrum, the user voice pattern, the one or more first topics, the stored trust score, and the updated trust score under a user ID associated with the user.

13. The computer system of claim 8 , wherein the stored trust score and the updated trust score are further determined by a Dirichlet distribution model having parameters obtained by training the Dirichlet distribution model with training samples of callers.

14. The computer system of claim 8 , further comprising:

requesting the user to opt in for a service;

based on determining the user refused to opt in for the service, setting the stored trust score to a default value; and

based on determining the stored trust score is set to the default value, causing the computing device of the receiver of the call to display that the user refused to opt in for the service, wherein the user is the caller.

15. A computer program product for identifying a spam call, the computer program product comprising: one or more computer-readable storage medium and program instructions stored on at least one of the one or more storage medium, the program instructions executable by a processor, the program instructions comprising:

program instructions to capture an audio data associated with a voice of a user;

program instructions to determine a user frequency spectrum and a user voice pattern associated with the voice of the user;

program instructions to identify, using natural language processing (NLP), one or more first topics communicated by the user in a previous conversation including the user as a caller;

based on detecting a call from the caller to a receiver, program instructions to identify the caller as the user based on matching the user frequency spectrum to a caller frequency spectrum determined from a current conversation in the call and the user voice pattern to a caller voice pattern determined from the current conversation in the call;

responsive to identifying the caller as the user, program instructions to display on a computing device of the receiver, a stored trust score associated with the caller, wherein the stored trust score indicates to the receiver whether the caller is trustworthy based on a previous call between the caller and a different receiver, wherein the previous call includes the previous conversation; and

program instructions to determine an updated trust score for the caller relative to the stored trust score based on comparing the one or more first topics communicated by the caller in the previous conversation to one or more second topics communicated by the caller in the current conversation with the receiver to identify differences between the one or more first topics and the one or more second topics communicated by a same person, wherein the updated trust score is determined based on a plurality of flag counters, wherein at least one of the plurality of flag counters are generated based on a discrepancy by the caller in a user name pronunciation as compared to a stored user name pronunciation.

16. The computer program product of claim 15 , wherein the user frequency spectrum and the caller frequency spectrum are determined by a non-Bayesian Gaussian mixture model.

17. The computer program product of claim 15 , wherein the user voice pattern and the caller voice pattern are determined by a Mel Frequency Cepstral Coefficients algorithm.

18. The computer program product of claim 15 , wherein the program instructions to identify, using NLP, the one or more first topics associated with the user from the previous conversation including the user further comprises:

program instructions to determine the one or more first topics by an automatic text summarizer and a Latent Dirichlet Allocation model.

19. The computer program product of claim 15 , further comprising:

program instructions to store the user frequency spectrum, the user voice pattern, the one or more first topics, the stored trust score, and the updated trust score under a user ID associated with the user.

20. The computer program product of claim 15 , wherein the stored trust score and the updated trust score are further determined by a Dirichlet distribution model having parameters obtained by program instructions to train the Dirichlet distribution model with training samples of callers.

Assignments (1)

Continuity (1)

References Cited (26)