A Text2SQL intelligent query and answer method and system based on LLM and vector model

By adopting an interactive joint model based on LLM and vector model in the Text2SQL question and answer system, the existing system's problem of accurately understanding user intentions and generating correct SQL statements in complex data tables is solved, achieving higher accuracy and efficiency.

CN119149575BActive Publication Date: 2025-03-28TURING AI INST NANJING CO LTD
2 Cites 1 Cited by

Patent Information

Application Number
CN202411657542.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-11-20
Publication Date
2025-03-28
Estimated Expiration
2044-11-20

AI Technical Summary

Technical Problem

The existing Text2SQL Q&A system is difficult to accurately understand user intentions in complex data tables and generate correct SQL statements, and the model has problems with input noise and high computing resource requirements in practical applications.

Method used

Using an interactive joint model based on LLM and vector model, through multi-dimensional data representation, task decoupling, model multi-task synchronization fine-tuning and link construction, an intelligent Text2SQL query question and answer chain is realized, including multiple stages of collaboration such as user data table recall, time extraction, SQL generation and execution, and answer generation.

Benefits of technology

The accuracy of vector model positioning and most relevant Few-shot recalls is improved, the complexity of Text2SQL tasks is optimized, and the accuracy and efficiency of the model in practical applications is improved.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present invention discloses a Text2SQL intelligent query question-answering method and system based on LLM and vector model, which aims to build an interactive joint model with LLM as the core and vector model as the auxiliary, and realizes intelligent Text2SQL query question-answering chain through multi-dimensional data representation, task decoupling, model multi-task synchronous fine-tuning and link construction, including SQL generation and execution, answer generation and other stages of collaboration, so as to convert the query requirements expressed by users in natural language into professional MySQL query statements. The present invention disassembles complex Text2SQL tasks into multiple relatively simple tasks through task decoupling design and link construction, optimizes task node scheduling and performs fault tolerance processing for model generation anomalies through link construction; at the same time, to assist user inspection, the present invention also displays the results of each subtask on the front end, and makes the subtask results editable, so that the final results can be traced.
Need to check novelty before this filing date? Find Prior Art