Schema Graph Table Linking for Natural Language Database Queries

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional database systems are inflexible and inefficient in handling large amounts of data, requiring rigid structured query formats and excessive computational resources, and often fail to parse relevant information due to excessive data volumes.

Innovation Solution

A schema linking system that converts natural language queries to structured database queries by generating a schema graph, identifying relevant tables, and using a machine learning model to optimize the query process, reducing the amount of data processed and improving computational efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If conventional systems use rigid structured query language (SQL) format for database queries, then query execution can be performed on large amounts of data, but the system becomes inflexible and difficult to operate

Engineering Contradiction:

Improvequery flexibilityVSAvoidquery format rigidity

Core Design Contradiction:

Ease of operationVSAdaptability or versatility

Solution Approach 1:

The patent introduces a natural language processing intermediary layer between the user and the database query system. This intermediary translates flexible natural language queries into structured SQL queries, allowing users to benefit from both the ease of natural language input and the precision of structured query execution.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system changes the parameter of query input from rigid structured formats to flexible natural language formats. By transforming the input parameter type, the system maintains operational ease while preserving the ability to execute precise database queries through automated translation.

Inventive Principle:

Principle #35Parameter changes

2Reliability

If conventional systems process queries across many different tables in large databases, then complete information can be retrieved, but excessive computational resources are required

Engineering Contradiction:

Improveinformation retrieval completenessVSAvoidcomputational resource consumption

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent extracts and identifies only the relevant tables needed to answer the natural language query, rather than processing all tables in the database. This extraction of necessary components reduces computational resource consumption while maintaining information retrieval completeness.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The system segments the database processing task by dividing it into table identification and query execution phases. By first identifying relevant tables and then processing only those specific tables, the system reduces overall computational load while ensuring complete information retrieval from necessary sources.

Inventive Principle:

Principle #1Segmentation

3Adaptability or versatility

If machine learning models are used to formulate structured database queries, then query flexibility improves, but resource costs associated with making large quantities of calls to machine learning models increase

Engineering Contradiction:

Improvequery formulation flexibilityVSAvoidmachine learning model call cost

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The patent performs preliminary action by pre-identifying relevant tables before making machine learning model calls for query formulation. This preliminary table identification step reduces the scope of subsequent machine learning processing, thereby reducing the number and cost of model calls while maintaining query flexibility.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12499098B2Generating a schema graph of sub-tables in a database for queries using a large language model

Publication Date: 2025.12.16 ADOBE INC

AI summary

The present disclosure relates to systems, non-transitory computer-readable media, and methods for linking a database schema to a natural language query. In particular, in some embodiments, the disclosed systems determine, from tables in a database schema, a subset of tables relevant to a natural language query by comparing embeddings for the tables in the database schema and embeddings for the natural language query. Additionally, in some implementations, the disclosed systems select, from a schema graph comprising nodes that represent the tables in the database schema, an additional table along a path between a pair of nodes representing a pair of tables from the subset of tables. Moreover, in some embodiments, the disclosed systems determine a set of relevant tables by appending the additional table to the subset of tables. Furthermore, in some implementations, the disclosed systems generate, from the set of relevant tables, a response for the natural language query.