Natural language question generation

Patent metadata

Field	Value
Publication number	US-12456015-B1
Application number	US-202318193693-A
Country	US
Kind code	B1
Filing date	Mar 31, 2023
Priority date	Mar 31, 2023
Publication date	Oct 28, 2025
Grant date	Oct 28, 2025

Abstract

Techniques for generating a natural language prompt to further a goal of a dialog, are described. During a dialog, the system receives one or more user inputs including a user question, a user response to the question, and a request to generate a further question following the response. The system determines ASR output data corresponding to the user inputs, and determines dialog history data of the dialog. Using the ASR output data and the dialog history data, the system determines a category and an explanation of relevance corresponding to the category. Using the ASR output data, the dialog history, the category, and the explanation, the system determines the further question to be output to the user.

First claim

What is claimed is:

1. A computer-implemented method comprising:
receiving first input data representing a first question corresponding to a first dialog in a natural language;
receiving second input data representing a first response to the first question;
receiving third input data corresponding to a request to generate a second question following the first response;
using a first machine learning (ML) component configured to generate data associated with a predicted question of a dialog, processing the first input data, the second input data, and the third input data to determine:
a first category corresponding to the second question, and
a first explanation of relevance of the first category;
processing, using a second ML component, the first input data, the second input data, the third input data, the first category, and the first explanation to generate the second question; and
outputting the second question.

2. The computer-implemented method of claim 1 , further comprising:
receiving fourth input data representing a third question corresponding to a second dialog in the natural language;
receiving fifth input data representing a second response to the third question;
processing, by a first preliminary ML component, the fourth input data and the fifth input data to determine:
a second category corresponding to a fourth question following the second response, and
a second explanation of relevance of the second category;
determining a first comparison of the second category and a training category corresponding to the third question;
determining a second comparison of the second explanation and a training natural explanation corresponding to the third question; and
based on the first comparison and the second comparison, determining a third ML component.

3. The computer-implemented method of claim 2 , further comprising:
after determining the third ML component, processing, using the third ML component, the fourth input data, the fifth input data, the second category, and the second explanation to generate the third question;
determining a third comparison of the third question and a training question; and
based on the third comparison, determining a fourth ML component, wherein the first ML component and the second ML component correspond to the fourth ML component.

4. The computer-implemented method of claim 1 , further comprising:
prior to receiving the first input data, receiving a transcript of a description of a first event associated with the first dialog;
using a third ML component, processing the transcript to generate a third question associated with the first event; and
outputting the third question.

5. A computer-implemented method comprising:
receiving first input data corresponding to a first natural language question;
receiving second input data corresponding to first natural language response to the first natural language question;
using a first machine learning (ML) component configured to generate data associated with a predicted question of a dialog, processing the first input data and the second input data to determine:
a first category corresponding to a second natural language question following the first natural language response, and
a first explanation of relevance of the first category;
using the first input data, the second input data, the first category, and the first explanation, generating the second natural language question; and
outputting the second natural language question.

6. The computer-implemented method of claim 5 , further comprising:
prior to receiving the first input data, receiving third input data corresponding to a natural language description of a first event;
using a second ML component, processing the third input data to generate a third natural language question associated with the first event; and
outputting the third natural language question.

7. The computer-implemented method of claim 5 , wherein:
the first input data and the second input data are associated with a first dialog involving a first user and a second user,
the first natural language question corresponds to a question asked by the first user to the second user,
the first natural language response corresponds to a response of the second user to the question asked by the first user, and
the second natural language question is configured to be addressed to the second user.

8. The computer-implemented method of claim 5 , wherein:
generating the second natural language question comprises processing the first input data, the second input data, the first category, and the first explanation using a second ML component.

9. The computer-implemented method of claim 5 , further comprising:
determining third input data corresponding to a third natural language question previously presented, wherein:
determining the first category is further based on processing the third input data,
determining the first explanation is further based on processing the third input data, and
determining the second natural language question is further based on processing the third input data.

10. The computer-implemented method of claim 5 , further comprising:
after outputting the second natural language question, receiving third input data corresponding to a second natural language response to the second natural language question;
receiving fourth input data corresponding to the second natural language question;
using the first ML component, processing the third input data and the fourth input data to determine:
a second category corresponding to a third natural language question following the second natural language response, and
a second explanation of relevance of the second category;
processing the third input data, the fourth input data, the second category, and the second explanation to generate the third natural language question;
determining that the third natural language question is semantically similar to the second natural language question; and
ceasing processing with respect to the third natural language question.

11. The computer-implemented method of claim 5 , further comprising:
after outputting the second natural language question, receiving third input data corresponding to a request to output an explanation of relevance of the second natural language question; and
based on the third input data corresponding to the request to output the explanation of relevance of the second natural language question, outputting the first explanation.

12. The computer-implemented method of claim 5 , wherein:
the first input data and the second input data are associated with a first dialog, the first dialog corresponding to a first goal to be achieved during the first dialog, and
the method further comprises:
using the first ML component, processing the first input data, the second input data, the first category, and the first explanation to generate the second natural language question, wherein the second natural language question is configured to result in third input data associated with the first goal.

13. A computing system comprising:
at least one processor; and
at least one memory comprising instructions that, when executed by the at least one processor, cause the computing system to:
receive first input data corresponding to a first natural language question;
receive second input data corresponding to a first natural language response to the first natural language question;
using a first machine learning (ML) component configured to generate data associated with a predicted question of a dialog, process the first input data and the second input data to determine:
a

Assignees

• Amazon Tech Inc

Inventors

• Kumar Anoop
• Fung Yi
• Natarajan Premkumar
• Galstyan Aram
• Ji Heng

Classifications

• G06F16/3329

  Natural language query formulation · CPC title

• G06F40/40

  Processing or translation of natural language (natural language analysis G06F40/20; semantic analysis G06F40/30) · CPC title

• G06F40/30 Primary

  Semantic analysis · CPC title