Efficiently processing query workloads with natural language statements and native database commands

Patent metadata

Field	Value
Publication number	US-12430333-B2
Application number	US-202418438224-A
Country	US
Kind code	B2
Filing date	Feb 9, 2024
Priority date	Feb 9, 2024
Publication date	Sep 30, 2025
Grant date	Sep 30, 2025

Abstract

A database query processing service is provided for efficiently processing query workloads with natural language statements and native database commands. The database query processing service may receive some database queries that do not contain a natural language marker and process these database queries without using large language models to generate replacement query content. The database query processing service may also receive other database queries that do contain the natural language marker and process the other database queries using large language model(s) to generate replacement query content or leverage replacement query content already generated by the large language model(s). The replacement query content is checked to ensure the content is natively valid for the content to retrieve data from database structures referenced in the content. The database query processing service may use the natively valid replacement query content to cause execution of operations responsive to the other database queries.

First claim

What is claimed is:

1. A computer-implemented method comprising:
receiving, by a database query processing service operating on one or more computing devices, a first database query comprising one or more commands and one or more conditions that are natively valid for a database query language to retrieve data from one or more database structures that are referenced in the first database query without using a large language model to generate natively valid replacement database query content, wherein the first database query does not contain a natural language marker;
causing execution, by the database query processing service, of the first database query to retrieve data from one or more database structures without using a large language model to generate natively valid replacement database query content between receiving the first database query and execution of the first database query;
receiving, by the database query processing service, a second database query;
determining, by the database query processing service, that the second database query contains the natural language marker and natural language text;
based at least in part on determining that the second database query contains the natural language marker and the natural language text, prompting a particular large language model for natively valid replacement database query content based at least in part on the natural language text in the second database query;
based at least in part on the prompting of the particular large language model, receiving particular database query content generated at least in part by the particular large language model;
determining whether the particular database query content is natively valid for the database query language to retrieve data from one or more particular database structures that are referenced in the particular database query content;
based at least in part on determining that the particular database query content is natively valid for the database query language to retrieve data from the one or more particular database structures that are referenced in the particular database query content, causing execution, by the database query processing service, of an operation responsive to the second database query using the particular database query content.

2. The computer-implemented method of claim 1 , wherein causing execution, by the database query processing service, of the operation comprises causing execution, by the database query processing service, of the particular database query content to retrieve particular data from the one or more particular database structures, and including the particular data in a response to the second database query.

3. The computer-implemented method of claim 1 , wherein causing execution, by the database query processing service, of the operation comprises providing, by the database query processing service in response to the second database query, the particular database query content and an option to proceed with execution of the particular database query content.

4. The computer-implemented method of claim 1 , wherein causing execution, by the database query processing service, of the operation comprises:
causing execution, by the database query processing service, of the particular database query content to retrieve particular data from the one or more particular database structures;
prompting a large language model to generate a natural language result explaining the particular data retrieved;
providing, by the database query processing service in response to the second database query, a natural language response based at least in part on the natural language result.

5. The computer-implemented method of claim 1 , further comprising prompting a large language model to generate a natural language result explaining logic of the particular database query content; wherein causing execution, by the database query processing service, of the operation comprises providing, by the database query processing service in response to the second database query, a natural language response based at least in part on the natural language result.

6. The computer-implemented method of claim 1 , further comprising providing, to the particular large language model, metadata identifying a plurality of data structures including the one or more particular data structures, wherein the metadata is associated with prompting the particular large language model for natively valid database query content based at least in part on the natural language text in the second database query.

7. The computer-implemented method of claim 6 , wherein the metadata is stored in association with a profile of a plurality of profiles, wherein other metadata is stored in association with at least one other profile of the plurality of profiles, wherein the metadata identifies at least one data structure not identified by the other metadata; the computer-implemented method further comprising:
before receiving the first database query or the second database query, receiving a request to set the profile as an active profile in a session of the database query processing service;
wherein the first database query and the second database query are received in the session, and wherein the profile remains as the active profile during execution of the first database query and execution of the operation responsive to the second database query in the session.

8. The computer-implemented method of claim 1 , further comprising providing, to the particular large language model, one or more example natively valid database queries that are not responsive to the second database query; wherein prompting the particular large language model for natively valid database query content based at least in part on the natural language text in the second database query uses the one or more example natively valid database queries that are not responsive to the second database query.

9. The computer-implemented method of claim 1 , further comprising:
receiving a selection of the particular large language model from a plurality of available large language models; and
storing credentials for accessing the particular large language model;
wherein prompting the particular large language model for natively valid replacement database query content based at least in part on the natural language text in the second database query uses one or more large language model instructions that are unique to the particular large language model of the plurality of large language models.

10. The computer-implemented method of claim 1 , further comprising, in a session with the particular large language model, prompting the particular large language model to generate a natural language result, wherein the particular large language model is configured to use a prompt history in the session to generate the natural language result, and wherein the natural language result is based at least in part on information provided in one or more past prompts to the particular large language model.

11. The computer-implemented method of claim 1 , wherein the natural language text is a first natural language text, further comprising:
storing the particular database query content in association with the first natural language text;
receiving, by the database query processing service, a third database query that includes the natural language marker and a second natural language text;
determining that the second natural language text is semantically equivalent to the first natural language text;
based at least in part on determining that the second natural language text is semantically equivalent to the first natural language text, causing execution, by the datab

Assignees

• Oracle Int Corp

Inventors

• Jain Sanket
• Rajamani Kumar

Classifications

• G06F40/279

  Recognition of textual entities · CPC title

• G06F40/30

  Semantic analysis · CPC title

• G06F40/186 Primary

  Templates · CPC title

• G06F16/24542 Primary

  Plan optimisation · CPC title