Granted Patent B1

US 11,250,876 · App. 16/707,345 · Granted Feb 15, 2022

Inventors: Connor Warren McCloskey (Bloomington, IL); Divya Pratap Singh Bhati (Arlington Heights, IL); Donna Gerig (Bloomington, IL)

Assignee: STATE FARM MUTUAL AUTOMOBILE INSURANCE COMPANY

G10L25/63G06N20/00G10L15/26H04M3/5183G06Q40/08

Abstract

Claims (51)

1. A computer implemented method for confidential sentiment analysis, comprising:

receiving call data of a call in an interaction recording system of a call center recorder, wherein the interaction recording system is located behind a firewall of an internal network sub-environment,

within the internal network sub-environment:

(i) storing the call data in an electronic database, wherein the call data includes interaction metadata,

(ii) generating a speech-to-text transcript by analyzing call audio in the call data, wherein the speech-to-text transcript corresponds to words spoken by one or more callers during the call,

(iii) generating an anonymized transcript corresponding to the speech-to-text transcript by identifying, removing, replacing, or obscuring one or more alpha-numeric personally identifiable information, and

(iv) generating a sentiment score by analyzing the anonymized transcript using a sentiment analysis service.

2. The method of claim 1 , wherein the interaction recording system is located behind the firewall of the internal network sub-environment includes collecting a consent of the caller to store the call data.

3. The method of claim 1 , wherein storing the call data includes storing caller identification information.

4. The method of claim 1 , wherein generating the anonymized transcript by identifying, removing, replacing, or obscuring one or more alpha-numeric personally identifiable information is based on matching regular expression patterns against the speech-to-text transcript.

5. The method of claim 1 , wherein generating the anonymized transcript by identifying, removing, replacing, or obscuring one or more alpha-numeric personally identifiable information is based on matching keywords in the speech-to-text transcript to one or more corpora of words.

6. The method of claim 1 , wherein generating the anonymized transcript by identifying, removing, replacing, or obscuring one or more alpha-numeric personally identifiable information is based on analyzing the speech-to-text transcript using a trained machine learning model.

7. The method of claim 1 , wherein generating the sentiment score by analyzing the anonymized transcript using the sentiment analysis service includes generating a time series wherein each time step corresponds to a time in the anonymized transcript, and each time step is associated with a sentiment score, wherein the sentiment score indicates the sentiment at the respective time step.

8. The method of claim 1 , wherein generating the sentiment score by analyzing the anonymized transcript using the sentiment analysis service includes generating an intra-call sentiment score.

9. A confidential sentiment analysis computing system, comprising

one or more processors, and

a memory including computer executable instructions that, when executed by the one or more processors, cause the computing system to:

receive call data of a call in an interaction recording system of a call center recorder, wherein the interaction recording system is located behind a firewall of an internal network sub-environment,

within the internal network sub-environment:

(i) store the call data in an electronic database, wherein the call data includes interaction metadata,

(ii) generate a speech-to-text transcript by analyzing call audio in the call data, wherein the speech-to-text transcript corresponds to words spoken by one or more callers during the call,

(iii) generate an anonymized transcript corresponding to the speech-to-text transcript by identifying, removing, replacing, or obscuring one or more alpha-numeric personally identifiable information, and

(iv) generate a sentiment score by analyzing the anonymized transcript using a sentiment analysis service.

10. The computing system of claim 9 , the memory containing further instructions that, when executed by the one or more processors, cause the computing system to:

collect a consent of the caller to store the call data.

11. The computing system of claim 9 , the memory containing further instructions that, when executed by the one or more processors, cause the computing system to:

generate the anonymized transcript by matching regular expression patterns against the speech-to-text transcript.

12. The computing system of claim 9 , the memory containing further instructions that, when executed by the one or more processors, cause the computing system to:

generate the anonymized transcript by matching keywords in the speech-to-text transcript to one or more corpora of words.

13. The computing system of claim 9 , the memory containing further instructions that, when executed by the one or more processors, cause the computing system to:

generate the anonymized transcript by analyzing the speech-to-text transcript using a trained machine learning model.

14. The computing system of claim 9 , the memory containing further instructions that, when executed by the one or more processors, cause the computing system to:

generate a time series wherein each time step corresponds to a time in the anonymized transcript, and each time step is associated with a sentiment score, wherein the sentiment score indicates the sentiment at the respective time step.

15. The computing system of claim 9 , the memory containing further instructions that, when executed by the one or more processors, cause the computing system to:

generating an intra-call sentiment score.

16. A non-transitory computer readable medium containing program instructions that when executed, cause a computer system to:

receive call data of a call in an interaction recording system of a call center recorder, wherein the interaction recording system is located behind a firewall of an internal network sub-environment,

within the internal network sub-environment:

(i) store the call data in an electronic database, wherein the call data includes interaction metadata,

(ii) generate a speech-to-text transcript by analyzing call audio in the call data, wherein the speech-to-text transcript corresponds to words spoken by one or more callers during the call,

(iii) generate an anonymized transcript corresponding to the speech-to-text transcript by identifying, removing, replacing, or obscuring one or more alpha-numeric personally identifiable information, and

(iv) generate a sentiment score by analyzing the anonymized transcript using a sentiment analysis service.

17. The non-transitory computer readable medium of claim 16 , including further program instructions that when executed, cause a computer system to:

collect a consent of the caller to store the call data.

18. The non-transitory computer readable medium of claim 16 ,

including further program instructions that when executed, cause a computer system to:

generate the anonymized transcript by matching regular expression patterns against the speech-to-text transcript.

19. The non-transitory computer readable medium of claim 16 , including further program instructions that when executed, cause a computer system to:

generate the anonymized transcript by matching keywords in the speech-to-text transcript to one or more corpora of words.

20. The non-transitory computer readable medium of claim 16 , including further program instructions that when executed, cause a computer system to:

generate the anonymized transcript by analyzing the speech-to-text transcript using a trained machine learning model.

Assignments (1)

Cited By (8)

US 12,190,906 US 12,217,752 US 12,223,087 US 12,282,880 US 12,375,605 US 12,393,719 US 12,446,837 US 12,651,596