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.
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
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.
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.
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.
Smart Images

Figure CN121960501B_ABST
Abstract
Citation Information
Patent Citations
Railway traffic field semantic recognition method and device based on few sample data and medium
CN121747577A