Granted Patent B2

US 12,387,059 · App. 17/554,606 · Granted Aug 12, 2025

Inventors: Akshay Aravindakshan Thoniparambil (Bangalore, IN); Naman Gupta (Jaipur, IN); Manish Agarwal (Bangalore, IN); Sourav Choudhary (Samastipur, IN)

Assignee: Capital One Services, LLC

G06F40/58G06F16/3344G06F16/93G06F40/216G06N3/08

Abstract

Claims (70)

1. A method, comprising:

receiving, by an application executing on a processor, input specifying a text statement type;

determining, by the application based on the text statement type, a plurality of heuristics for identifying text statements of the text statement type in a plurality of text transcripts;

determining, by the application based on a first heuristic of the plurality of heuristics, a first text statement of the text statement type;

generating, by the application based on a clustering algorithm and the first text statement, a plurality of additional text statements of the text statement type;

assigning, by a classification model, a classification tag to each of the plurality of additional text statements, the classification tag indicating that each of the plurality of additional text statements is of the text statement type, wherein the classification model is trained based on a training dataset comprising a plurality of training terms, wherein one or more of the plurality of training terms is masked in the training dataset based on a respective frequency each training term appears among a plurality of terms in the plurality of text transcripts and whether the respective frequency exceeds a threshold;

receiving, by the application based on the first text statement and the plurality of additional text statements, a first text transcript of the plurality of text transcripts;

identifying, by the application based on a second heuristic of the plurality of heuristics, a first text transcript statement in the first text transcript, wherein the first text transcript statement is of the text statement type;

generating a graphical indication that the first text transcript statement is of the text statement type;

displaying at least the first text transcript statement and the graphical indication on a display;

the method further comprising:

determining, by the application based on a third heuristic of the plurality of heuristics, a plurality of seed words related to the text statement type; and

generating, by the application based on the clustering algorithm and the plurality of seed words, the plurality of additional text statements of the text statement type; and

the method further comprising, prior to receiving the input:

receiving, by the application, the training dataset comprising the plurality of training terms;

determining, by the application, the respective frequency each training term appears among the plurality of terms in the plurality of text transcripts;

classifying, by the application, a first subset of the plurality of training terms as being associated with the text statement type based on the respective frequencies of the first subset exceeding the threshold;

classifying, by the application, a second subset of the plurality of training terms as not being associated with the text statement type based on the respective frequencies of the second subset not exceeding the threshold; and

masking the first subset of the plurality of training terms from the training dataset.

2. The method of claim 1 , wherein the first text transcript statement is further identified based on one or more of: (i) a length of the first text transcript statement, (ii) whether a predetermined phrase is present in the first text transcript statement, (iii) whether the first text transcript statement is associated with a customer or a customer support agent, (iv) a sentiment score computed for the first text transcript statement, and (v) whether a term of the first text transcript statement is associated with a negative sentiment.

3. The method of claim 1 , wherein the text statement types comprise one or more of: (i) a reason a customer associated with the text transcript contacted a customer support agent, (ii) a call resolution associated with the customer support agent, (iii) a personal story associated with the customer, and (iv) a negative statement made by the customer.

4. The method of claim 1 , wherein the first text transcript is identified based on one or more statements of the first text transcript matching the first text statement or at least one of the plurality of additional text statements, wherein the graphical indication comprises one or more of: (i) a symbol, (ii) a highlighting effect applied to the first text transcript statement, (iii) a bold effect applied to the first text transcript statement, or (iv) an italics effect applied to the first text transcript statement.

5. The method of claim 1 , wherein the text statement type is defined as being separate from a statement in the plurality of text transcripts.

6. A non-transitory computer-readable storage medium, the computer-readable storage medium including instructions that when executed by a processor, cause the processor to:

receive, by an application, input specifying a text statement type;

determine, by the application based on the text statement type, a plurality of heuristics for identifying statements of the text statement type in a plurality of text transcripts;

determine, by the application based on a first heuristic of the plurality of heuristics, a first text statement of the text statement type;

generate, by the application based on a clustering algorithm and the first text statement, a plurality of additional text statements of the text statement type;

assign, by a classification model, a classification tag to each of the plurality of additional text statements, the classification tag indicating that each of the plurality of additional text statements is of the text statement type, wherein the classification model is trained based on a training dataset comprising a plurality of training terms, wherein one or more of the plurality of training terms is masked in the training dataset based on a respective frequency each training term appears among a plurality of terms in the plurality of text transcripts and whether the respective frequency exceeds a threshold;

receive, by the application based on the first text statement and the plurality of additional text statements, a first text transcript of the plurality of text transcripts;

identify, by the application based on a second heuristic of the plurality of heuristics, a first text transcript statement in the first text transcript, wherein the first text transcript statement is of the text statement type;

generate a graphical indication that the first text transcript statement is of the text statement type;

display at least the first text transcript statement and the graphical indication on a display;

the instructions further causing the processor to:

determine, by the application based on a third heuristic of the plurality of heuristics, a plurality of seed words related to the text statement type; and

generate, by the application based on the clustering algorithm and the plurality of seed words, the plurality of additional text statements of the text statement type; and

the instructions further causing the processor to, prior to receiving the input:

receive, by the application, the training dataset comprising the plurality of training terms;

determine, by the application, the respective frequency each training term appears among a plurality of terms in the plurality of text transcripts;

classify, by the application, a first subset of the plurality of training terms as being associated with the text statement type based on the respective frequencies of the first subset exceeding the threshold;

classify, by the application, a second subset of the plurality of training terms as not being associated with the text statement type based on the respective frequencies of the second subset not exceeding the threshold; and

mask the first subset of the plurality of training terms from the training dataset.

7. The computer-readable storage medium of claim 6 , wherein the first text transcript statement is further identified based on one or more of: (i) a length of the first text transcript statement, (ii) whether a predetermined phrase is present in the first text transcript statement, (iii) whether the first text transcript statement is associated with a customer or a customer support agent, (iv) a sentiment score computed for the first text transcript statement, and (v) whether a term of the first text transcript statement is associated with a negative sentiment.

8. The computer-readable storage medium of claim 6 , wherein the text statement types comprise one or more of: (i) a reason a customer associated with the text transcript contacted a customer support agent, (ii) a call resolution associated with the customer support agent, (iii) a personal story associated with the customer, and (iv) a negative statement made by the customer.

9. The computer-readable storage medium of claim 6 , wherein the first text transcript is identified based on one or more statements of the first text transcript matching the first text statement or at least one of the plurality of additional text statements, wherein the graphical indication comprises one or more of: (i) a symbol, (ii) a highlighting effect applied to the first text transcript statement, (iii) a bold effect applied to the first text transcript statement, or (iv) an italics effect applied to the first text transcript statement.

10. The computer-readable storage medium of claim 6 , wherein the text statement type is defined as being separate from a statement in the plurality of text transcripts.

11. A computing apparatus comprising:

a processor; and

a memory storing instructions that, when executed by the processor, cause the processor to:

receive, by an application, input specifying a text statement type;

determine, by the application based on the text statement type, a plurality of heuristics for identifying statements of the text statement type in a plurality of text transcripts;

determine, by the application based on a first heuristic of the plurality of heuristics, a first text statement of the text statement type;

generate, by the application based on a clustering algorithm and the first text statement, a plurality of additional text statements of the text statement type;

assign, by a classification model, a classification tag to each of the plurality of additional text statements, the classification tag indicating that each of the plurality of additional text statements is of the text statement type, wherein the classification model is trained based on a training dataset comprising a plurality of training terms, wherein one or more of the plurality of training terms is masked in the training dataset based on a respective frequency each training term appears among a plurality of terms in the plurality of text transcripts and whether the respective frequency exceeds a threshold;

receive, by the application based on the first text statement and the plurality of additional text statements, a first text transcript of the plurality of text transcripts;

identify, by the application based on a second heuristic of the plurality of heuristics, a first text transcript statement in the first text transcript, wherein the first text transcript statement is of the text statement type;

generate a graphical indication that the first text transcript statement is of the text statement type;

display at least the first text transcript statement and the graphical indication on a display;

the instructions further causing the processor to:

determine, by the application based on a third heuristic of the plurality of heuristics, a plurality of seed words related to the text statement type; and

generate, by the application based on the clustering algorithm and the plurality of seed words, the plurality of additional text statements of the text statement type; and

the instructions further causing the processor to, prior to receiving the input:

receive, by the application, the training dataset comprising the plurality of training terms;

determine, by the application, a respective frequency each training term appears among the plurality of terms in the plurality of text transcripts;

classify, by the application, a first subset of the plurality of training terms as being associated with the text statement type based on the respective frequencies of the first subset exceeding the threshold;

classify, by the application, a second subset of the plurality of training terms as not being associated with the text statement type based on the respective frequencies of the second subset not exceeding the threshold; and

mask the first subset of the plurality of training terms from the training dataset.

12. The computing apparatus of claim 11 , wherein the first text transcript statement is further identified based on one or more of: (i) a length of the first text transcript statement, (ii) whether a predetermined phrase is present in the first text transcript statement, (iii) whether the first text transcript statement is associated with a customer or a customer support agent, (iv) a sentiment score computed for the first text transcript statement, and (v) whether a term of the first text transcript statement is associated with a negative sentiment.

13. The computing apparatus of claim 11 , wherein the text statement types comprise one or more of: (i) a reason a customer associated with the text transcript contacted a customer support agent, (ii) a call resolution associated with the customer support agent, (iii) a personal story associated with the customer, and (iv) a negative statement made by the customer, wherein the first text transcript is identified based on one or more statements of the first text transcript matching the first text statement or at least one of the plurality of additional text statements, wherein the graphical indication comprises one or more of: (i) a symbol, (ii) a highlighting effect applied to the first text transcript statement, (iii) a bold effect applied to the first text transcript statement, or (iv) an italics effect applied to the first text transcript statement.

14. The computing apparatus of claim 11 , wherein the text statement type is defined as being separate from a statement in the plurality of text transcripts.

Assignments (1)

Continuity (1)

References Cited (35)