Method and system for performing table question-answering tasks while preserving data security

Patent metadata

Field	Value
Publication number	US-2025190617-A1
Application number	US-202318535428-A
Country	US
Kind code	A1
Filing date	Dec 11, 2023
Priority date	Dec 11, 2023
Publication date	Jun 12, 2025
Grant date	—

Abstract

A method for using a large language model to perform complex table question-answering tasks while preserving data security and privacy with respect to the underlying data included in a table is provided. The method includes: receiving a user query that relates to a table; generating, based on the user query, a prompt that includes a role, at least one instruction, schema, and a question; transmitting the prompt to an external processor that does not have access to the underlying data included in the table; receiving, from the external processor, a set of executable code which has been generated by the external processor in response to the prompt and which is applicable to the table; generating an answer to the user query by executing the set of executable code; and transmitting, to the user, the answer to the user query.

First claim

What is claimed is:

1 . A method for performing a question-answering task relating to a table while preserving data security and privacy with respect to underlying data included in the table, the method being implemented by a first processor, the method comprising:
receiving, by the first processor from a user, a user query that relates to the table; generating, by the first processor based on the user query, a prompt that includes a role, at least one instruction, schema, and a question; transmitting, by the first processor to a second processor, the prompt; receiving, by the first processor from the second processor, a first set of executable code that is applicable to the table, the first set of executable code having been generated by the second processor in response to the prompt; generating, by the first processor, an answer to the user query by executing the first set of executable code; and transmitting, by the first processor to the user, the answer to the user query.

2 . The method of claim 1 , wherein the role includes an explanation that the second processor is an artificial intelligence assistant that is designed to answer questions from tables by generating executable code and that the second processor is provided with information that relates to columns included in the table but not provided with the underlying data included in the table.

3 . The method of claim 1 , wherein the at least one instruction includes information indicating a type of the executable code to be generated and at least one suggestion that relates to how to overcome a potential determination by the second processor that the question is not answerable.

4 . The method of claim 1 , wherein the schema includes information that indicates a number of columns included in the table and column header information for each of the columns included in the table.

5 . The method of claim 1 , wherein the question includes information that relates to the user query.

6 . The method of claim 1 , further comprising determining whether the answer to the user query corresponds to a successful retrieval of information from the table.

7 . The method of claim 6 , wherein the successful retrieval includes at least one from among a text entry from the table, an aggregated numerical value, and a list of entries from the table.

8 . The method of claim 1 , further comprising:
determining whether the answer to the user query corresponds to a failure; and when a determination is made that the answer to the user query corresponds to the failure, generating an error message that includes at least one from among information indicating that a value error has been made, information indicating that an index error has been made, information indicating that a null answer that provides no identifiable answer and no executable code has been generated, and information indicating that the second processor has generated a comment corresponding to an inability of the second processor to generate a response to the prompt.

9 . The method of claim 8 , wherein when the determination is made that the answer corresponds to the failure, the method further comprises:
transmitting the error message to the second processor; receiving, from the second processor, a second set of executable code that is applicable to the table; and generating an updated answer by executing the second set of executable code.

10 . A computing apparatus for performing a question-answering task relating to a table while preserving data security and privacy with respect to underlying data included in the table, the computing apparatus comprising:
a first processor; a memory; and a communication interface coupled to each of the first processor and the memory, wherein the first processor is configured to:
receive, from a user via the communication interface, a user query that relates to the table;
generate, based on the user query, a prompt that includes a role, at least one instruction, schema, and a question;
transmit, via the communication interface to an external processor that is outside of the computing apparatus, the prompt;
receive, from the external processor via the communication interface, a first set of executable code that is applicable to the table, the first set of executable code having been generated by the external processor in response to the prompt;
generate an answer to the user query by executing the first set of executable code; and
transmit, to the user via the communication interface, the answer to the user query.

11 . The computing apparatus of claim 10 , wherein the role includes an explanation that the external processor is an artificial intelligence assistant that is designed to answer questions from tables by generating executable code and that the external processor is provided with information that relates to columns included in the table but not provided with the underlying data included in the table.

12 . The computing apparatus of claim 10 , wherein the at least one instruction includes information indicating a type of the executable code to be generated and at least one suggestion that relates to how to overcome a potential determination by the external processor that the question is not answerable.

13 . The computing apparatus of claim 10 , wherein the schema includes information that indicates a number of columns included in the table and column header information for each of the columns included in the table.

14 . The computing apparatus of claim 10 , wherein the question includes information that relates to the user query.

15 . The computing apparatus of claim 10 , wherein the first processor is further configured to determine whether the answer to the user query corresponds to a successful retrieval of information from the table.

16 . The computing apparatus of claim 15 , wherein the successful retrieval includes at least one from among a text entry from the table, an aggregated numerical value, and a list of entries from the table.

17 . The computing apparatus of claim 10 , wherein the first processor is further configured to:
determine whether the answer to the user query corresponds to a failure; and when a determination is made that the answer to the user query corresponds to the failure, generate an error message that includes at least one from among information indicating that a value error has been made, information indicating that an index error has been made, information indicating that a null answer that provides no identifiable answer and no executable code has been generated, and information indicating that the external processor has generated a comment corresponding to an inability of the second processor to generate a response to the prompt.

18 . The computing apparatus of claim 17 , wherein when the determination is made that the answer corresponds to the failure, the first processor is further configured to:
transmit the error message to the external processor; receive, from the external processor, a second set of executable code that is applicable to the table; and generate an updated answer by executing the second set of executable code.

19 . A non-transitory computer readable storage medium storing instructions for performing a question-answering task relating to a table while preserving data security and privacy with respect to underlying data included in the table, the storage medium comprising a first set of executable code which, when executed by a first processor, causes the first processor to:
receiv

Assignees

• Jpmorgan Chase Bank Na

Inventors

• Watson William
• Cho Naan
• Balch Tucker Richard
• Veloso Manuela

Classifications

• G06F21/6227

  where protection concerns the structure of data, e.g. records, types, queries · CPC title

• G06F21/6245 Primary

  Protecting personal data, e.g. for financial or medical purposes · CPC title