Dfa and large language model based hybrid architecture semantic repair method and device

By using a hybrid architecture of DFA and a large language model, the problem of matching colloquial text with standardized terminology in the power sector was solved, achieving efficient and accurate semantic repair and improving the operational efficiency and decision-making accuracy of power business systems.

CN121960501BActive Publication Date: 2026-07-21POWERCHINA JIANGXI ELECTRIC POWER ENGINEERING CO LTD
View PDF 1 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
POWERCHINA JIANGXI ELECTRIC POWER ENGINEERING CO LTD
Filing Date
2026-03-31
Publication Date
2026-07-21

AI Technical Summary

Technical Problem

Existing technologies cannot effectively solve the problem of matching colloquial texts with standardized terminology in the power sector, resulting in knowledge retrieval failures, low accuracy in data structuring, and an inability to meet the demands for high-concurrency processing efficiency and semantic accuracy.

Method used

A hybrid architecture based on DFA and a large language model is adopted. Non-standard terminology candidate fragments are located through millisecond-level full-scale scanning. The semantic coupling probability is calculated by combining a large model with thinking chain reasoning ability, the repair path is determined and semantic repair is performed to ensure the structural integrity and semantic accuracy of power professional terms.

Benefits of technology

It has achieved efficient conversion of colloquial power language texts into standardized terminology, improved the knowledge recall purity of the RAG system, reduced decision bias, and ensured the operational efficiency and decision accuracy of the power business system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121960501B_ABST
    Figure CN121960501B_ABST
Patent Text Reader

Abstract

The application discloses a kind of based on DFA and big language model's hybrid architecture semantic repair method and device, it is related to semantic repair technical field, the method includes: receiving original power text sequence, loading professional term benchmark library and based on DFA carries out full amount scanning to output candidate anchor point set;Lock candidate anchor point set and extract the context feature window of each anchor point, calculate the semantic coupling probability of each anchor point and the context in context feature window;Determine the repair path between each anchor point and the context in context feature window based on semantic coupling probability, and based on repair path, each anchor point and context are repaired to obtain corresponding standardized term fragment Semantics;Standardized term fragment is integrated, and the complete standardized text sequence corresponding to original power text sequence is formed.The application solves the problem of low accuracy and efficiency in converting spoken language text into standardized terms in the prior art.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Railway traffic field semantic recognition method and device based on few sample data and medium

    CN121747577A