A question and answer analysis method, device and medium

By using intent recognition and semantic parsing, combined with a knowledge base and indicator management system, the system automatically completes the dimensional values ​​of the query elements in user questions, solving the problem of low accuracy caused by incomplete data elements in business intelligence systems and improving the efficiency and accuracy of data queries.

CN122242528APending Publication Date: 2026-06-19SHENZHEN LEXIN SOFTWARE TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
SHENZHEN LEXIN SOFTWARE TECH CO LTD
Filing Date
2026-03-17
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

Existing business intelligence systems suffer from low accuracy in intent recognition and data querying due to missing or incomplete data element dimension values ​​in user natural language input, resulting in the inability to automatically complete data and leading to erroneous data or errors.

Method used

By using intent recognition and semantic parsing, the system identifies user intent using a pre-trained natural language processing model and a multi-classifier. Combined with a knowledge base and an indicator management system, it completes the query element dimension values ​​in the user's question and generates a complete SQL query statement.

Benefits of technology

It improves the accuracy of intent recognition and the efficiency of data retrieval, enhances the system's tolerance for incomplete queries, ensures that users obtain reliable business data, and optimizes the interactive experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122242528A_ABST
    Figure CN122242528A_ABST
Patent Text Reader

Abstract

This invention relates to the field of semantic analysis technology, and discloses a question-and-answer analysis method, device, and medium, comprising: when the intent type is indicator consultation, using the retrieved indicator metadata information as the question answer corresponding to the user's question; when the intent type is multidimensional data analysis, assigning attributes based on the numerical values ​​corresponding to the question query elements to divide the question query elements into a first query element and a second query element; when the element dimension value does not exist in the dimension enumeration value set or the element dimension value is missing, completing the element dimension value based on the dimension enumeration value set, and determining the dimension value in the second query element as the default value; generating an SQL query statement corresponding to the user's question based on the completed first query element, the second query element determined to have the default value, and the metadata, executing the SQL query statement to obtain the question answer corresponding to the user's question. This invention can improve the accuracy of question-and-answer analysis.
Need to check novelty before this filing date? Find Prior Art