System and method of sentiment modeling and application to determine optimized agent action

• US 10,735,586 B1
• Filed: 12/13/2019
• Issued: 08/04/2020
• Est. Priority Date: 10/15/2019
• Status: Active Grant

• Current Assignee: Verint Americas Incorporated (Verint Systems Incorporated)
• Original Assignee: Verint Americas Incorporated (Verint Systems Incorporated)
• Inventors: Johnston, Michael
• Primary Examiner(s): Deane, Jr., William J
• Application Number:
  US16/713,813
• Time in Patent Office: 235 Days
• Field of Search: 37926501-26514, 37926501-2651, 379309
• CPC Class Codes: G06F 40/216 using statistical methods
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06Q 10/10 Office automation; Time management
• CPC Class Codes: G06Q 30/01 Customer relationship services
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: H04M 2201/40 using speech recognition speech recognition per se G10L15/00
• CPC Class Codes: H04M 2203/401 Performance feedback
• CPC Class Codes: H04M 3/5175 Call or contact centers supervision arrangements
• CPC Class Codes: H04M 3/5183 Call or contact centers with computer-telephony arrangements

Abstract

The present invention is a system and method of continuous sentiment tracking and the determination of optimized agent actions through the training of sentiment models and applying the sentiment models to new incoming interactions. The system receives conversations comprising incoming interactions and agent actions and determines customer sentiment on a micro-interaction level for each incoming interaction. Based on interaction types, the system correlates the determined sentiment with the agent action received prior to the sentiment determination to create and train sentiment models. Sentiment models include agent action recommendations for a desired sentiment outcome. Once trained, the sentiment models can be applied to new incoming interactions to provide CSRs with actions that will yield a desired sentiment outcome.

Claims

1. A method for processing conversations to provide continuous sentiment analysis and optimized agent actions, comprising:

receiving an incoming interaction, wherein the incoming interaction is part of a conversation;

performing a MAS analysis of the incoming interaction using a metadata analytics service (MAS) software module on a MAS;

generating interaction metadata for the incoming interaction based on the MAS analysis;

passing the incoming communication and the interaction metadata associated with the incoming interaction to a sentiment analysis engine (SAE);

performing a SAE analysis of the incoming interaction using a SAE software module based on the interaction metadata and a set of sentiment criteria;

assigning a sentiment to the incoming interaction based on the SAE analysis;

receiving an agent action in response to the incoming interaction, wherein the agent action is part of the conversation;

performing an agent MAS analysis of the agent action using the MAS software module on the MAS;

generating interaction metadata for the agent action based on the agent MAS analysis;

continuing to receive and analyze additional incoming interactions and additional agent actions, wherein the additional incoming interactions and the additional agent actions are part of the conversation, until the conversation is ended;

receiving the conversation at a model analysis engine (MAE), when the conversation is ended;

performing a model analysis of the agent actions from the conversation along with the sentiment assigned to the incoming interactions from the conversation to correlate agent actions and corresponding sentiment to at least one of a plurality of interaction types;

creating at least one sentiment model for each interaction type represented in the conversation that does not already have a sentiment model based on the model analysis, wherein the sentiment model includes at least one optimized agent action and one of the interaction types;

updating the sentiment models for each interaction type represented in the conversation based on the model analysis;

receiving, at a recommendation analysis engine (RAE), a new incoming interaction along with associated sentiment for the new incoming interaction;

receiving the at least one sentiment model at the RAE;

applying the at least one sentiment model to the new incoming interaction to determine an optimized agent action; and

displaying to a customer service representative the sentiment for the new incoming interaction and the determined optimized agent action.

2. The method of claim 1, wherein the incoming interaction and the agent action includes relation metadata, wherein the relation metadata includes a set of information defining how the incoming interaction and the agent action are related to other incoming interactions and other agent actions of the conversation.

3. The method of claim 1, wherein the interaction metadata includes a determination of at least one interaction type for the incoming interaction.

4. The method of claim 1, the method further comprising determining an end of the conversation by analyzing the incoming interaction and the agent action.

5. The method of claim 1, the method further comprising determining at least one interaction type for the incoming interaction based on an analysis of the incoming interaction and the interaction metadata.

6. The method of claim 1, the method further comprising determining at least one conversation type for the conversation based on an analysis of the incoming interactions for the conversation and the agent actions for the conversation.

7. The method of claim 1, wherein the sentiment model only includes agent actions for the interaction type that results in a positive sentiment.

8. The method of claim 1, wherein the sentiment model only includes agent actions for the interaction type that results in a negative sentiment.

9. The method of claim 1, wherein the sentiment is one of positive, negative, or neutral.

10. The method of claim 9, wherein the at least one optimized agent action for each sentiment model is the agent action most likely to yield a specific sentiment for the interaction type.

11. An automated computer system for automatedly providing continuous sentiment analysis and optimized agent actions, comprising:

a processor;

a display with a user interface to display a sentiment for a new incoming interaction and at least one determined optimized agent action to a customer service representative (CSR); and

a non-transitory computer readable medium programmed with computer readable code that upon execution by the processor causes the processor to;

receive an incoming interaction, wherein the incoming interaction is part of a conversation,

perform a MAS analysis of the incoming interaction using a metadata analytics service (MAS) software module on a MAS, generate interaction metadata for the incoming interaction based on the MAS analysis,

pass the incoming communication and the interaction metadata associated with the incoming interaction to a sentiment analysis engine (SAE), perform a SAE analysis of the incoming interaction using a SAE software module based on the interaction metadata and a set of sentiment criteria, assign a sentiment to the incoming interaction based on the SAE analysis, receive an agent action in response to the incoming interaction, wherein the agent action is part of the conversation,

perform an agent MAS analysis of the agent action using the MAS software module on the MAS,

generate interaction metadata for the agent action based on the agent MAS analysis,

continue to receive and analyze additional incoming interactions and additional agent actions, wherein the additional incoming interactions and the additional agent actions are part of the conversation, until the conversation is ended,

receive the conversation at a model analysis engine (MAE), when the conversation is ended,

perform a model analysis of the agent actions from the conversation along with the sentiment assigned to the incoming interactions from the conversation to correlate agent actions and corresponding sentiment to at least one of a plurality of interaction types,

create at least one sentiment model for each interaction type represented in the conversation that does not already have a sentiment model based on the model analysis, wherein the sentiment model includes at least one optimized agent action and one of the interaction types, update the sentiment models for each interaction type represented in the conversation based on the model analysis, receive, at a recommendation analysis engine (RAE), a new incoming interaction along with associated sentiment for the new incoming interaction, receive the at least one sentiment model at the RAE,

apply the at least one sentiment model to the new incoming interaction to determine an optimized agent action, and

display to a customer service representative the sentiment for the new incoming interaction and the determined optimized agent action.

12. The system of claim 11, wherein the incoming interaction and the agent action includes relation metadata, wherein the relation metadata includes a set of information defining how the incoming interaction and the agent action are related to other incoming interactions and other agent actions of the conversation.

13. The system of claim 11, wherein the interaction metadata includes a determination of at least one interaction type for the incoming interaction.

14. The system of claim 11, wherein execution of the non-transitory computer readable medium further causes the processor to determine an end of the conversation by analyzing the incoming interaction and the agent action.

15. The system of claim 11, wherein execution of the non-transitory computer readable medium further causes the processor to determine at least one interaction type for the incoming interaction based on an analysis of the incoming interaction and the interaction metadata.

16. The system of claim 11, wherein execution of the non-transitory computer readable medium further causes the processor to determine at least one conversation type for the conversation based on an analysis of the incoming interactions for the conversation and the agent actions for the conversation.

17. The system of claim 11, wherein the sentiment model only includes agent actions for the interaction type that results in a positive sentiment.

18. The system of claim 11, wherein the sentiment model only includes agent actions for the interaction type that results in a negative sentiment.

19. The system of claim 11, wherein the at least one optimized agent action for each sentiment model is the agent action most likely to yield a specific sentiment for the interaction type.

20. A non-transitory computer readable medium programmed with computer readable code that upon execution by a processor causes the processor to execute a method for automatedly providing continuous sentiment analysis and optimized agent actions, comprising:

receiving an incoming interaction, wherein the incoming interaction is part of a conversation;

performing a MAS analysis of the incoming interaction using a metadata analytics service (MAS) software module on a MAS;

generating interaction metadata for the incoming interaction based on the MAS analysis;

passing the incoming communication and the interaction metadata associated with the incoming interaction to a sentiment analysis engine (SAE);

performing a SAE analysis of the incoming interaction using a SAE software module based on the interaction metadata and a set of sentiment criteria;

assigning a sentiment to the incoming interaction based on the SAE analysis;

receiving an agent action in response to the incoming interaction, wherein the agent action is part of the conversation;

performing an agent MAS analysis of the agent action using the MAS software module on the MAS;

generating interaction metadata for the agent action based on the agent MAS analysis;

continuing to receive and analyze additional incoming interactions and additional agent actions, wherein the additional incoming interactions and the additional agent actions are part of the conversation, until the conversation is ended;

receiving the conversation at a model analysis engine (MAE), when the conversation is ended;

performing a model analysis of the agent actions from the conversation along with the sentiment assigned to the incoming interactions from the conversation to correlate agent actions and corresponding sentiment to at least one of a plurality of interaction types;

creating at least one sentiment model for each interaction type represented in the conversation that does not already have a sentiment model based on the model analysis, wherein the sentiment model includes at least one optimized agent action and one of the interaction types;

updating the sentiment models for each interaction type represented in the conversation based on the model analysis;

receiving, at a recommendation analysis engine (RAE), a new incoming interaction along with associated sentiment for the new incoming interaction;

receiving the at least one sentiment model at the RAE;

applying the at least one sentiment model to the new incoming interaction to determine an optimized agent action; and

displaying to a customer service representative the sentiment for the new incoming interaction and the determined optimized agent action.