Learning and automating agent actions

• US 10,026,092 B2
• Filed: 12/09/2016
• Issued: 07/17/2018
• Est. Priority Date: 12/09/2016
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Heater, John Dolan, Rohatgi, Abhishek, Negrean, Flaviu Gelu
• Primary Examiner(s): GAUTHIER, GERALD
• Application Number:
  US15/373,879
• Publication Number:
  US 20180165691A1
• Time in Patent Office: 585 Days
• Field of Search: 370352, 379 671, 37921801, 37926503, 704 9, 704239, 704257, 705 5, 705 732, 705 144, 705 1454, 705 1464, 705 64, 705304, 709200, 709204, 715738
• CPC Class Codes: G06F 40/205 Parsing
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06Q 30/016 After-sales
• CPC Class Codes: H04L 51/02 using automatic reactions or user delegation, e.g. automatic replies or chatbot-generated messages
• CPC Class Codes: H04M 2203/402 Agent or workforce management
• CPC Class Codes: H04M 3/51 Centralised call answering arrangements requiring operator intervention , e.g. call or contact centers for telemarketing

Abstract

An automated system processes customer requests and interactions with a customer service representative. During such interactions, the representative'"'"'s actions at a computer interface are monitored, and the dialog between the customer and representative are recorded. Based on the dialog and the actions, a script is created or updated for a given request, where the script encompasses a dialog tree and actions relating to the customer'"'"'s account. When a subsequent customer submits the same or a comparable request, an automated agent utilizes the script to handle the request. Using the script, the automated agent performs a dialog with the customer, accesses the customer account, and updates the account in accordance with the request.

Claims

1. A method of processing customer requests, comprising:

parsing a request by a customer to identify key words of the request;

parsing a dialog between the customer and an agent;

monitoring operations taken by the agent at a computer terminal during occurrence of the dialog;

identifying an intent based on at least one of the key words, dialog and monitored operations, the intent indicating an objective of the request;

generating a script based on the dialog and the monitored operations, the script including

     1) a dialog tree between a subsequent customer and an automated agent, and

     2) instructions for the agent to perform at least one action on the subsequent customer'"'"'s account, the at least one action being coupled to at least one node of the dialog tree; and

updating a database to incorporate the script and the intent.

2. The method of claim 1, wherein the request is a first request by a first customer, and further comprising:

parsing a second request by a second customer to identify key words of the second request;

comparing the key words against the database to locate an existing script for the second request;

in response to locating the existing script for the second request in the database;

retrieving the existing script from the database, and

executing the existing script by

     1) conducting an automated dialog with the second customer based on the script, and

     2) performing at least one action on the second customer'"'"'s account based on the script; and

in response to failing to locate the existing script for the second request;

forwarding the second request to the agent,

parsing a dialog between the second customer and an agent;

monitoring operations taken by the agent at the computer terminal during occurrence of the dialog;

identifying a second intent based on at least one of the key words, dialog and monitored operations associated with the second customer, the second intent indicating an objective of the second request;

generating a second script based on the dialog and the monitored operations, the second script including

     1) a second dialog tree between the subsequent customer and the automated agent, and

     2) instructions to perform at least one action on the subsequent customer'"'"'s account, the at least one action being coupled to at least one node of the second dialog tree; and

updating the database to incorporate the second script and the second intent.

3. The method of claim 1, wherein the request is a first request by a first customer, and further comprising:

parsing a second request by a second customer to identify key words of the second request;

parsing a dialog between the second customer and the agent;

monitoring operations taken by the agent at a computer terminal during occurrence of the dialog;

identifying an intent for the second request based on at least one of the key words, dialog, and monitored operations, the intent for the second request matching the intent for the first request; and

updating the database to modify the script based on the dialog between the second customer and the agent and the monitored operations.

4. The method of claim 1, wherein updating the database includes adding at least one entry to the database, the at least one entry including the intent and the script.

5. The method of claim 1, wherein updating the database includes:

locating an existing script in the database, the existing script being associated with the intent; and

modifying the existing script, based on the script, to generate a modified script, the modified script including portions of both the script and the existing script.

6. The method of claim 1, wherein the operations taken by the agent include at least one of accessing the customer'"'"'s account at an account database, retrieving information from the customer'"'"'s account, and modifying the customer'"'"'s account.

7. The method of claim 1, wherein monitoring the operations includes monitoring use of a user input device by the agent, the user input device including at least one of a keyboard, mouse, touchscreen, and voice interface.

8. The method of claim 1, further comprising associating each of the operations with a respective portion of the dialog based on a time at which each of the operations occurs relative to the dialog.

9. The method of claim 1, further comprising transcribing spoken utterances by at least one of the customer and the agent to obtain at least one of the request and the dialog.

10. A method of processing customer requests, comprising:

parsing a request by a customer to identify key words of the request;

comparing the key words against a database to locate a script for the request;

in response to locating a script for the request in the database;

retrieving the script from the database, and

executing the script by

     1) conducting an automated dialog with the customer based on the script, and

     2) performing at least one action on the customer'"'"'s account based on the script; and

in response to failing to locate a script for the request;

forwarding the request to an agent,

parsing a dialog between the customer and an agent;

monitoring operations taken by the agent at a computer terminal during occurrence of the dialog;

identifying an intent based on at least one of the key words, dialog and monitored operations, the intent indicating an objective of the request;

generating a script based on the dialog and the monitored operations, the script including

     1) a dialog tree between a subsequent customer and an automated agent, and

     2) instructions to perform at least one action on the subsequent customer'"'"'s account, the at least one action being coupled to at least one node of the dialog tree; and

updating a database to incorporate the script and the intent.

11. The method of claim 10, further comprising:

comparing the key words against the database to determine the intent; and

searching the database, based on the intent, to locate a script associated with the intent.

12. The method of claim 10, wherein executing the script includes accessing the customer'"'"'s account via a virtual desktop environment, the at least one action including an operation within the virtual desktop environment.

13. The method of claim 10, wherein parsing the dialog includes performing semantic parsing on the dialog to generate a semantic map of the dialog.

14. The method of claim 13, wherein generating the script includes generating the dialog tree based on the semantic map of the dialog.

15. A method of processing customer requests, comprising:

parsing a plurality of customer requests to identify key words of the requests;

parsing a plurality of dialogs between the customer and an agent occurring in response to the plurality of customer requests;

monitoring operations taken by the agents at computer terminals during occurrence of the plurality of dialogs;

identifying an intent based on at least one of the key words, dialog and monitored operations, the intent indicating a common objective of the plurality of requests;

generating a script based on the plurality of dialogs and the monitored operations, the script including

     1) a dialog tree between a subsequent customer and an automated agent, and

     2) instructions to perform at least one action on the subsequent customer'"'"'s account, the at least one action being coupled to at least one node of the dialog tree; and

updating a database to incorporate the script and the intent.

16. The method of claim 15, wherein parsing the plurality of dialogs includes performing semantic parsing on the plurality of dialogs to generate a semantic map of the plurality of dialogs.

17. The method of claim 16, wherein generating the script includes generating the dialog tree based on the semantic map of the plurality of dialogs.

18. The method of claim 15, further comprising associating the at least one action with a respective portion of the dialog based on a time at which each of the operations occurs relative to the dialog.

19. The method of claim 15, further comprising transcribing spoken utterances by at least one of the customer and the agent to obtain at least one of the requests and the plurality of dialogs.