Method and system for code generation by large language models

Patent metadata

Field	Value
Publication number	US-12585439-B2
Application number	US-202318205719-A
Country	US
Kind code	B2
Filing date	Jun 5, 2023
Priority date	Mar 22, 2023
Publication date	Mar 24, 2026
Grant date	Mar 24, 2026

Abstract

A method for using a large language model (LLM) to generate executable code in a manner that preserves privacy and confidentiality of proprietary data is provided. The method includes: receiving information that relates to a summarization of a data table; defining a set of rules for facilitating a generation of executable code by an LLM; receiving an inquiry from a user; inputting each of the summarization information, the set of rules, and the inquiry into the LLM; receiving, in response to the input, a set of executable code that is generated by the LLM; and executing the set of executable code in order to generate an output. The summarization information may be received from an external source that has access to proprietary data included in the data table, and the summarization information may be structured so as to preserve a privacy of the proprietary data.

First claim

What is claimed is:

1 . A method for generating executable code, the method being implemented by at least one processor, the method comprising:
receiving, by the at least one processor and from an external source, first information that relates to a summarization of a first data table, wherein the external source has access to proprietary data that is included in the first data table, and wherein the summarization includes a textual description of each of a plurality of sections of the first data table; defining, by the at least one processor, a first set of rules for facilitating a generation of executable code by a language model, wherein the first set of rules include at least one rule that pertains to a programming language of the executable code, a library to be used in conjunction with the executable code, and information relating to a format of an output that reflects a user intent; receiving, by the at least one processor, an inquiry from a user; revising, by the at least one processor, the inquiry received based on the at least one rule included in the first set of rules; inputting, by the at least one processor, each of the first information, the first set of rules, and the revised inquiry into the language model, wherein the inputted first information omits the proprietary data included in the first data table; receiving, by the at least one processor in response to the inputting, a first set of executable code that is generated by the language model; and executing, by the at least one processor, the first set of executable code in order to generate the output that does not include the proprietary data included in the first data table but is rooted from the proprietary data included in the first data table.

2 . The method of claim 1 , wherein the first information is structured so as to preserve a privacy of the proprietary data.

3 . The method of claim 1 , further comprising inspecting, by a validator, the first set of executable code by examining an abstract syntax tree associated with the first set of executable code.

4 . The method of claim 1 , wherein the output includes at least one from among a second data table, a sunburst chart, a bar chart, a pie chart, an aggregated answer, a web page, a user interface, a slide presentation, a dashboard, and a document.

5 . The method of claim 1 , wherein the executing of the first set of executable code causes a generation of at least one from among an intermediate representation of the output and a data structure that corresponds to the output.

6 . The method of claim 1 , further comprising displaying the output on a display via a graphical user interface (GUI).

7 . A computing apparatus for generating executable code, the computing apparatus comprising:
a processor; a memory; a display; and a communication interface coupled to each of the processor, the memory, and the display, wherein the processor is configured to:
receive, via the communication interface and from an external source, first information that relates to a summarization of a first data table, wherein the external source has access to proprietary data that is included in the first data table, and wherein the summarization includes a textual description of each of a plurality of sections of the first data table;
define a first set of rules for facilitating a generation of executable code by a language model, wherein the first set of rules include at least one rule that pertains to a programming language of the executable code, a library to be used in conjunction with the executable code, and information relating to a format of an output that reflects a user intent;
receive, via the communication interface, an inquiry from a user;
revise the inquiry received based on the at least one rule included in the first set of rules;
input each of the first information, the first set of rules, and the revised inquiry into the language model, wherein the inputted first information omits the proprietary data included in the first data table;
receive, in response to the input via the communication interface, a first set of executable code that is generated by the language model; and
execute the first set of executable code in order to generate the output that does not include the proprietary data included in the first data table but is rooted from the proprietary data included in the first data table.

8 . The computing apparatus of claim 7 , wherein the first information is structured so as to preserve a privacy of the proprietary data.

9 . The computing apparatus of claim 7 , wherein the processor is further configured to inspect, via a validator, the first set of executable code by examining an abstract syntax tree associated with the first set of executable code.

10 . The computing apparatus of claim 7 , wherein the output includes at least one from among a second data table, a sunburst chart, a bar chart, a pie chart, an aggregated answer, a web page, a user interface, a slide presentation, a dashboard, and a document.

11 . The computing apparatus of claim 7 , wherein the processor is further configured to receive, from the language model, at least one from among an intermediate representation of the output and a data structure that corresponds to the output.

12 . The computing apparatus of claim 7 , wherein the processor is further configured to cause the display to display the output via a graphical user interface (GUI).

13 . A non-transitory computer readable storage medium storing instructions for generating a first set of executable code, the storage medium comprising a second set of executable code which, when executed by a processor, causes the processor to:
receive, from an external source, first information that relates to a summarization of a first data table, wherein the external source has access to proprietary data that is included in the first data table, and wherein the summarization includes a textual description of each of a plurality of sections of the first data table; define a first set of rules for facilitating a generation of executable code by a language model, wherein the first set of rules include at least one rule that pertains to a programming language of the executable code, a library to be used in conjunction with the executable code, and information relating to a format of an output that reflects a user intent; receive an inquiry from a user; revise the inquiry received based on the at least one rule included in the first set of rules; input each of the first information, the first set of rules, and the revised inquiry into the language model, wherein the inputted first information omits the proprietary data included in the first data table; receive, in response to the input, the first set of executable code that is generated by the language model; and execute the first set of executable code in order to generate the output that does not include the proprietary data included in the first data table but is rooted from the proprietary data included in the first data table.

Assignees

• Jpmorgan Chase Bank Na

Inventors

• Watson William
• Cho Naan
• Veloso Manuela

Classifications

• G06F21/604

  Tools and structures for managing or administering access control systems · CPC title

• G06F8/35 Primary

  model driven · CPC title