An intelligent question and answer method, device and system based on a knowledge graph and a text block

By converting natural language questions into vector form and combining them with knowledge graphs and text block sets for matching, the problem of insufficient accuracy of intelligent question answering systems in the securities industry is solved, achieving deep semantic understanding and efficient knowledge support.

CN121808024BActive Publication Date: 2026-06-26HUAAN SECURITIES CO LTD
View PDF 3 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
HUAAN SECURITIES CO LTD
Filing Date
2026-03-09
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Existing intelligent question-answering systems struggle to accurately retrieve relevant knowledge and achieve deep semantic understanding in highly specialized and complex fields, such as the securities industry, resulting in insufficient accuracy.

Method used

We employ an intelligent question-answering method based on knowledge graphs and text blocks. By converting natural language questions into vector form, we use knowledge graphs and text block sets for matching, combined with a pre-trained question-answering model, to achieve deep semantic understanding and accurate retrieval.

Benefits of technology

It improves the accuracy and comprehensiveness of intelligent question-answering systems in professional fields, providing efficient, accurate, and comprehensive knowledge support to meet the business needs of the securities industry.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121808024B_ABST
    Figure CN121808024B_ABST
Patent Text Reader

Abstract

The application relates to the technical field of intelligent question answering, and discloses an intelligent question answering method, device and system based on a knowledge graph and a text block. The application first converts a user question into a vector question, and then matches the vector question with the knowledge graph and a text block set respectively. When the knowledge graph is matched, an initial subgraph is obtained first, and then the initial subgraph is matched with the user question. When the text block set is matched, a second relevant text block is obtained based on the matching of the question vector and the text block set, and then the first relevant text block and the second relevant text block are combined to obtain a target relevant text block. Finally, all the target relevant text blocks and relevant subgraphs are input into a question answering model, the text block provides detailed text basis for the model, and the subgraph presents entity association logic in a structured form, so that the question answering model can accurately retrieve relevant knowledge and achieve deep semantic understanding, and the accuracy of intelligent question answering is improved.
Need to check novelty before this filing date? Find Prior Art