Generating multi-order text query results utilizing a context orchestration engine

Patent metadata

Field	Value
Publication number	US-2024281621-A1
Application number	US-202318309496-A
Country	US
Kind code	A1
Filing date	Apr 28, 2023
Priority date	Feb 21, 2023
Publication date	Aug 22, 2024
Grant date	—

Abstract

The present disclosure is directed toward systems, methods, and non-transitory computer-readable media for generating responses to multi-order text queries using a context orchestration engine. For example, the disclosed systems generate context-defining query subcomponents from a multi-order text query, where the context-defining query subcomponents indicate contextual data sources pertaining to their respective portions of the multi-order text query. In addition, the disclosed systems provide or transmit the context-defining query subcomponents to a large language model for domain-specific computer code pertaining to each respective context-defining query subcomponent. The disclosed systems can further execute the generated computer code for each context-defining query subcomponent to access indicated contextual data sources for generating component-specific results. The disclosed systems can also generate a multi-order result to the multi-order text query from the component-specific results.

First claim

What is claimed is:

1 . A computer-implemented method comprising:
generating, from a multi-order text query received from a client device, a first context-defining query subcomponent indicating a first contextual data source for responding to the first context-defining query subcomponent; providing the first context-defining query subcomponent to a large language model for generating computer code that is specific to the first contextual data source and executable for responding to the first context-defining query subcomponent; executing the computer code from the large language model using the first contextual data source accessed over a computer network to generate a first result to the first context-defining query subcomponent of the multi-order text query; identifying a second result to a second context-defining query subcomponent generated from the multi-order text query, the second context-defining query subcomponent indicating a second contextual data source; and utilizing the first result and the second result to generate an aggregated result to the multi-order text query.

2 . The computer-implemented method of claim 1 , wherein generating the first context-defining query subcomponent further comprises providing the multi-order text query to the large language model to generate the first context-defining query subcomponent.

3 . The computer-implemented method of claim 1 , wherein generating the first context-defining query subcomponent comprises determining that the multi-order text query is a non-first order text query.

4 . The computer-implemented method of claim 1 , further comprises:
identifying a third result to a third context-defining query subcomponent generated from the multi-order text query that indicates the first contextual data source; and utilizing the first result, the second result, and the third result, to generate the aggregated result to the multi-order text query.

5 . The computer-implemented method of claim 1 , wherein providing the first context-defining query subcomponent to the large language model for generating computer code comprises transmitting the first context-defining query subcomponent to the large language model to cause the large language model to generate the computer code in a domain-specific computer language specific to the first contextual data source indicated by the first context-defining query subcomponent.

6 . The computer-implemented method of claim 1 , wherein generating the first result comprises at least one of a determination of the first context-defining query subcomponent or an action of the first context-defining query subcomponent.

7 . The computer-implemented method of claim 1 , wherein generating the first result comprises performing a first action, the first action comprising a component-specific action within the first contextual data source.

8 . The computer-implemented method of claim 1 , further comprising:
receiving, from the large language model, a request for sample data for generating the first result; and providing, to the large language model, the sample data for training the large language model to generate the first result corresponding to the first context-defining query subcomponent.

9 . The computer-implemented method of claim 1 , further comprising, in response to utilizing the first result and the second result to generate the aggregated result to the multi-order text query, providing a response to the client device that indicates the aggregated result.

10 . The computer-implemented method of claim 1 , wherein the first contextual data source comprises a first application within an organizational ecosystem.

11 . The computer-implemented method of claim 1 , wherein the second contextual data source comprises a third-party application not within an organizational ecosystem.

12 . A system comprising:
at least one processor; and a non-transitory computer readable medium comprising instructions that, when executed by the at least one processor, cause the system to:
generate, from a multi-order text query received from a client device, a first context-defining query subcomponent indicating a first contextual data source for responding to the first context-defining query subcomponent;
provide the first context-defining query subcomponent to a large language model for generating domain-specific computer code that is specific to the first contextual data source and executable for responding to the first context-defining query subcomponent;
execute the domain-specific computer code from the large language model using the first contextual data source accessed over a computer network to generate a first result to the first context-defining query subcomponent of the multi-order text query, the domain-specific computer code generates a component-specific result utilizing data stored at the first contextual data source;
identify a second result to a second context-defining query subcomponent generated from the multi-order text query, the second context-defining query subcomponent indicating a second contextual data source; and
utilize the first result and the second result to generate an aggregated result to the multi-order text query.

13 . The system of claim 12 , further comprising instructions that, when executed by the at least one processor, cause the system to:
determine that the multi-order text query is a non-first order text query; and identify a third result to a third context-defining query subcomponent generated from the multi-order text query that indicates the first contextual data source.

14 . The system of claim 13 , further comprising instructions that, when executed by the at least one processor, cause the system to utilize the first result, the second result, and the third result to generate the aggregated result to the multi-order text query.

15 . The system of claim 12 , further comprising instructions that, when executed by the at least one processor, cause the system to generate at least one of a determination related to the first context-defining query subcomponent or an action of the first context-defining query subcomponent.

16 . The system of claim 12 , further comprising instructions that, when executed by the at least one processor, cause the system to:
determine that the first context-defining query subcomponent is unanswerable; and send a request to the client device to provide additional context for the first context-defining query subcomponent, the additional context comprising an indication of an additional contextual data source.

17 . The system of claim 12 , further comprising instructions that, when executed by the at least one processor, cause the system to:
determine that the first context-defining query subcomponent is unanswerable in response to receiving, from the large language model, a request for sample data for generating a component-specific result corresponding to the first context-defining query subcomponent of the multi-order text query; and provide, to the large language model, contextual sample data for training the large language model to generate the component-specific result corresponding to the first context-defining query subcomponent of the multi-order text query.

18 . A non-transitory computer readable medium comprising instructions that, when executed by at least one processor, cause the at least one processor to:
generate, from a multi-order text query received from a client device, a first context-defining query subcomponent indicating a first contextual data source for re

Assignees

• Dropbox Inc

Inventors

• Janakiraman Rajkumar
• Kulkarni Ranjitha Gurunath

Classifications

• G06F40/47 Primary

  Machine-assisted translation, e.g. using translation memory · CPC title

• G06F40/51 Primary

  Translation evaluation · CPC title