Learning framework for processing communication session transcripts

Patent metadata

Field	Value
Publication number	US-12197865-B2
Application number	US-202117554143-A
Country	US
Kind code	B2
Filing date	Dec 17, 2021
Priority date	Dec 17, 2021
Publication date	Jan 14, 2025
Grant date	Jan 14, 2025

Abstract

Learning frameworks for processing text transcripts may include receiving, by an application, a query comprising a topic. A topic model may determine a plurality of subtopics based on the topic. The application may receive, from a database based on the topic and plurality of subtopics, a plurality of text transcripts. A sentiment model may compute, for each text transcript, a respective sentiment score based on a text of the respective text transcript. The application may determine, for each text transcript, a duration of a communication session associated with the respective text transcript. The application may compute, for each text transcript, a total score based on the sentiment score and the duration of the respective text transcript. The application may return, as responsive to the query, a subset of the plurality of text transcripts having a total score that exceeds a threshold.

First claim

What is claimed is:

1. A method, comprising:
receiving, by an application executing on a processor, a query comprising a topic;
determining, by a topic model executing on the processor, a plurality of subtopics based on the topic;
receiving, by the application from a database based on the topic and plurality of subtopics, a plurality of text transcripts, each transcript associated with a respective communication session;
computing, by a sentiment model executing on the processor for each text transcript, a respective sentiment score based on a text of the respective text transcript;
determining, by the application for each text transcript, a duration of the communication session associated with the respective text transcript;
computing, by the application for each text transcript, a total score based on the sentiment score and the duration of the respective communication session;
returning, by the application as responsive to the query, a subset of the plurality of text transcripts having a total score that exceeds a threshold;
extracting, by a key phrase model, a plurality of key phrases from each text transcript of the subset of the plurality of text transcripts;
determining, by the application and based on content of the plurality of key phrases, a system error associated with the application; and
transmitting, by the application, a notification including the system error to a computing device to address the system error.

2. The method of claim 1 , further comprising:
determining, by the application for each communication session, a first amount of time a customer engaged in conversation with an agent;
computing, by the application for each text transcript, a first time score based on the first amount of time;
determining, by the application for each communication session, a second amount of time a customer was on hold; and
computing, by the application for each text transcript, a second time score based on the second amount of time, wherein the total score is further based on the first and second time scores.

3. The method of claim 2 , wherein computing the total score comprises computing a sum of the sentiment score, the first time score, and the second time score.

4. The method of claim 1 , further comprising:
outputting, by the application, the plurality of key phrases for display.

5. The method of claim 1 ,
wherein the notification includes the topic determined from the topic model and one or more portions of the plurality of text transcripts that are associated with the topics.

6. The method of claim 4 , wherein the key phrase model is trained based on a plurality of training transcripts using unsupervised training, wherein the topic model is based on the key phrase model using semi-supervised training.

7. The method of claim 1 , wherein determining the plurality of subtopics is based on clustering the topic into a cluster and identifying the plurality of subtopics in the cluster; and
wherein clustering the topic is performed using an unsupervised training algorithm.

8. 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, a query comprising a topic;
determine, by a topic model, a plurality of subtopics based on the topic;
receive, by the application from a database based on the topic and plurality of subtopics, a plurality of text transcripts, each transcript associated with a respective communication session;
compute, by a sentiment model for each text transcript, a respective sentiment score based on a text of the respective text transcript;
determine, by the application for each text transcript, a duration of the communication session associated with the respective text transcript;
compute, by the application for each text transcript, a total score based on the sentiment score and the duration of the respective text transcript;
return, by the application as responsive to the query, a subset of the plurality of text transcripts having a total score that exceeds a threshold;
extract, by a key phrase model, a plurality of key phrases from each text transcript of the subset of the plurality of text transcripts;
determine, by the application and based on content of the plurality of key phrases, a system error associated with the application; and
transmit, by the application, a notification including the system error to a computing device to address the system error.

9. The computer-readable storage medium of claim 8 , wherein the instructions further cause the processor to:
determine, by the application for each communication session, a first amount of time a customer engaged in conversation with an agent;
compute, by the application for each text transcript, a first time score based on the first amount of time;
determine, by the application for each communication session, a second amount of time a customer was on hold; and
compute, by the application for each text transcript, a second time score based on the second amount of time, wherein the total score is further based on the first and second time scores.

10. The computer-readable storage medium of claim 9 , wherein compute the total score comprises computing a sum of the sentiment score, the first time score, and the second time score.

11. The computer-readable storage medium of claim 8 , wherein the instructions further configure the computer to:
output, by the application, the plurality of key phrases for display.

12. The computer-readable storage medium of claim 8 ,
wherein the notification includes the topic determined from the topic model and one or more portions of the plurality of text transcripts that are associated with the topics.

13. The computer-readable storage medium of claim 11 , wherein the key phrase model is trained based on a plurality of training transcripts use unsupervised training, wherein the topic model is based on the key phrase model using semi-supervised training.

14. The computer-readable storage medium of claim 8 , wherein determining the plurality of subtopics is based on clustering the topic into a cluster and identifying the plurality of subtopics in the cluster; and
wherein clustering the topic is performed using a dbscan clustering algorithm.

15. 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, a query comprising a topic;
determine, by a topic model, a plurality of subtopics based on the topic;
receive, by the application from a database based on the topic and plurality of subtopics, a plurality of text transcripts, each transcript associated with a respective communication session;
compute, by a sentiment model for each text transcript, a respective sentiment score based on a text of the respective text transcript;
determine, by the application for each text transcript, a duration of the communication session associated with the respective text transcript;
compute, by the application for each text transcript, a total score based on the sentiment score and the duration of the respective text transcript;
return, by the application as responsive to the query, a subset of the plurality of text transcripts having a total score that exceeds a threshold;
extract, by a key phrase model, a plurality of key phrases from each text transcript of the subset of the plurality of text transcripts;
determine, by the application and based on content of the plurality of key phrases, a system error associated with

Assignees

• Capital One Services Llc

Inventors

• Thoniparambil Akshay Aravindakshan
• Agarwal Manish
• Kedia Gaurav
• Bansiwala Aashish
• Ramamohan Deepak
• Choudhary Sourav

Classifications

• G06F40/35

  Discourse or dialogue representation · CPC title

• G06F16/3344

  using natural language analysis · CPC title

• G06F16/355

  Creation or modification of classes or clusters · CPC title

• G06F16/338

  Presentation of query results · CPC title

• G06F40/279

  Recognition of textual entities · CPC title