Automated service agents

• US 10,382,370 B1
• Filed: 08/11/2016
• Issued: 08/13/2019
• Est. Priority Date: 08/11/2016
• Status: Expired due to Fees

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Labarre, Thibaut, Keung, Hung Chiu Phillip, Lu, Yichao, Parvathaneni, Ranjit
• Primary Examiner(s): Alam, Uzma
• Application Number:
  US15/234,200
• Time in Patent Office: 1,097 Days
• CPC Class Codes: H04L 51/02 using automatic reactions or user delegation, e.g. automatic replies or chatbot-generated messages
• CPC Class Codes: H04L 51/04 Real-time or near real-time messaging, e.g. instant messaging [IM]
• CPC Class Codes: H04L 51/214 using selective forwarding

Abstract

Disclosed are various embodiments to deploy automated service agents, which can be implemented with an application executed on a computing device. The application can cause the computing device to process a transcript to identify an intended action of a service agent, wherein the transcript comprises a first set of text generated by the service agent and a second set of text generated by another party. The application can also cause the computing device, in response to an identification of the intended action from the transcript, to process the transcript to extract a minimum set of information required to create a workflow item that corresponds to the intended action. Moreover, the application can cause the computing device to send a request to a workflow management system to create the workflow item that corresponds to the intended action, wherein the request comprises an identifier for the intended action and the acquired information.

Claims

1. A system, comprising:

a computing device comprising a processor and a memory; and

an application comprising machine readable instructions stored in the memory that, when executed by the processor, cause the computing device to at least;

receive a transcript of a conversation that represents an interaction between a chat bot and a customer, wherein a portion of the interaction between the chat bot and the customer is generated with a recurrent neural network integrated with the chat bot;

determine that a portion of the transcript of the conversation between the chat bot and the customer matches a regular expression, wherein the regular expression is linked to an intended action to be performed by a service agent on behalf of the customer;

determine that the portion of the transcript is associated with clickstream data related to the intended action, the clickstream data comprising a plurality of records of sequences of interactions with a set of user interface elements by the service agent to perform the intended action;

extract from the transcript a minimum set of information required to create a workflow item corresponding to the intended action; and

send a request to a workflow management system to create the workflow item corresponding to the intended action, wherein the request comprises the minimum set of information.

2. The system of claim 1, wherein the application further comprises machine readable instructions stored in the memory that, when executed by the processor, cause the computing device to at least send to the chat bot a recommended response for the chat bot to provide to the customer, wherein the recommended response is based at least in part on the intended action.

3. The system of claim 1, wherein the transcript is generated from an audio file that has been processed with a speech-to-text conversion system.

4. The system of claim 1, wherein the intended action comprises at least one of:

a modification to an order, a cancellation of the order, or a cancellation of the user account.

5. A system, comprising:

a computing device comprising a processor and a memory; and

an application comprising machine readable instructions stored in the memory that, when executed by the processor, cause the computing device to at least;

compare a transcript with clickstream data to identify an intended action of a service agent by matching a portion of the transcript to a regular expression, wherein;

the transcript comprises a first set of text generated by the service agent and a second set of text generated by another party,

the clickstream data comprises a plurality of records of sequences of interactions with a set of user interface elements by the service agent to perform the intended action on behalf of the other party, and

wherein the regular expression is linked to the intended action of the service agent;

in response to an identification of the intended action from the transcript, process the transcript to extract a minimum set of information required to create a workflow item that corresponds to the intended action; and

send a request to a workflow management system to create the workflow item that corresponds to the intended action, wherein the request comprises an identifier for the intended action and the acquired information.

6. The system of claim 5, wherein the request to the workflow management system comprises a second request and the application further comprises machine readable instructions stored in the memory that, when executed by the processor, cause the computing device to at least:

determine that the transcript fails to include the minimum set of information required to create the workflow item; and

send a first request to the service agent to obtain a missing portion of the minimum set of information required to create the workflow item.

7. The system of claim 5, wherein the application further comprises machine readable instructions stored in the memory that, when executed by the processor, cause the computing device to at least perform a speech-to-text conversion on an audio stream to generate the transcript.

8. The system of claim 5, wherein the machine readable instructions that cause the computing device to at least process the transcript to identify the intended action of the service agent further comprise machine readable instructions that, when executed by the processor, cause the computing device to apply a random forest decision tree to the transcript to identify the intended action of the service agent.

9. The system of claim 5, wherein the application further comprises machine readable instructions stored in the memory that, when executed by the processor, cause the computing device to at least send to the service agent a recommended message for the service agent to provide to the user, wherein the recommended message is based at least in part on the intended action.

10. The system of claim 5, wherein the service agent comprises a chat bot and the first set of text is generated with a recurrent neural network trained with a plurality of transcripts.

11. The system of claim 5, wherein the intended action comprises a cancellation of an order.

12. The system of claim 5, wherein the intended action comprises a cancellation of a user account.

13. A method, comprising:

identifying, with a computing device, a section of a transcript of a conversation between a service agent and a user;

comparing, with the computing device the section of the transcript with clickstream data obtained from a plurality of interactions by the service agent with a service agent portal to determine that the section of the transcript corresponds to an intended action by the service agent on behalf of the user, the clickstream data comprising a plurality of records of sequences of interactions with a set of user interface elements of the service agent portal;

determining that the section of the transcript matches a regular expression linked to the intended action;

extracting, with the computing device, from the transcript a minimum set of information required to create the workflow item; and

sending, with the computing device, a request to a workflow management system to create a workflow item that corresponds to the intended action, wherein the request to create the workflow item comprises the minimum set of information.

14. The method of claim 13, further comprising performing, with the computing device, a speech-to-text conversion on audio data to generate the transcript.

15. The method of claim 13, further comprising generating, with the computing device, a response for the service agent based at least in part on the intended action.

16. The method of claim 13, wherein the service agent comprises a chat bot that uses a recurrent neural network to generate a response to a query from the user, wherein the transcript includes the response to the query.

17. The method of claim 13, wherein the intended action comprises a modification to an order.

18. The method of claim 13, wherein the intended action comprises a cancellation of an order.

19. The method of claim 13, wherein the intended action comprises a cancellation of the user account.