Conversational Data Table Querying With Partitioned LLM Mediation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Traditional methods of interacting with data tables face challenges such as inaccurate interpretation of complex natural language queries, lack of contextual understanding, data privacy issues, and incorrect outcomes due to reliance on Large Language Models (LLMs), especially when handling proprietary and confidential information.

Innovation Solution

A virtual assistant server that iteratively executes determined table operations on data tables based on natural language queries, using LLMs to generate responses while maintaining data privacy and security, by determining table intents and sub-intents, and orchestrating conversational interactions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If the entire data table is exposed to LLMs to answer user queries, then the language understanding and query handling capability is improved, but data privacy and security are compromised

Engineering Contradiction:

Improvequery handling capabilityVSAvoiddata privacy risk

Core Design Contradiction:

ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The patent segments the data table into multiple partitions and processes only the relevant partition identified by the user's natural language query. This selective processing approach maintains data privacy by limiting LLM exposure to only necessary data portions while preserving query handling capability.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary layer between the data table and LLM that identifies and retrieves only the relevant data partition needed to answer the query. This intermediary mechanism prevents direct exposure of entire data tables to LLMs, thereby protecting data privacy while maintaining effective query response.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If LLMs are used to process natural language queries, then language understanding is improved, but hallucination and bias issues arise leading to incorrect outcomes

Engineering Contradiction:

Improvelanguage understandingVSAvoidquery accuracy

Core Design Contradiction:

ReliabilityVSMeasurement precision

Solution Approach 1:

The patent implements a feedback mechanism where the system verifies LLM-generated responses against the actual data in the relevant partition. This feedback loop identifies and corrects hallucinations or biases by comparing model outputs with ground truth data, thereby improving query accuracy while maintaining language understanding capabilities.

Inventive Principle:

Principle #23Feedback

3Object-affected harmful factors

If traditional text-to-SQL techniques are used, then data privacy is maintained, but accurate interpretation of complex natural language queries fails

Engineering Contradiction:

Improvedata privacy protectionVSAvoidquery interpretation accuracy

Core Design Contradiction:

Object-affected harmful factorsVSReliability

Solution Approach 1:

The patent introduces an intermediary natural language processing layer that enhances traditional text-to-SQL approaches. This intermediary uses LLMs to improve understanding of complex queries while maintaining data privacy by processing only identified relevant partitions rather than exposing entire data tables.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent segments the query processing into multiple stages: natural language understanding, relevant partition identification, and targeted query execution. This segmentation allows complex natural language queries to be accurately interpreted while limiting data exposure to only necessary portions.

Inventive Principle:

Principle #1Segmentation

4Reliability

If LLMs process unstructured text, then language processing is improved, but handling of structured data becomes limited

Engineering Contradiction:

Improveunstructured text processingVSAvoidstructured data handling

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent merges the strengths of LLMs for unstructured text processing with traditional structured data query capabilities. By combining natural language understanding with partition-based structured data retrieval, the system achieves both improved language processing and effective structured data handling.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS20260030240A1Systems and methods for enabling conversational interaction with data tables

Publication Date: 2026.01.29 KORE AI INC

AI summary

A method for enabling users to conversationally interact with a data table comprises receiving by a virtual assistant server from a user device, a natural language query targeted at a data table. The virtual assistant server determines one or more table intents or one or more table sub-intents of the natural language query. The virtual assistant server determines one or more table operations to execute on the data table based on the determined table intents or table sub-intents. The virtual assistant server iteratively executes each of the determined table operations on the data table, where a final version of the data table results after completion of one or more of the iteratively executed determined table operations. Subsequently, the virtual assistant server determines a response to the natural language query based on the final version of the data table and outputs the determined response to the natural language query to the user device.