Complaint analysis method based on LLM large model RAG technology
By using LLM big model and RAG technology in the analysis of government complaint data, the subject information of the complaint work order is extracted and judged, and the misjudgment problem of similar or same complaint subjects in the existing technology is solved, and the accuracy of statistical data is achieved.
CN119719122AActive Publication Date: 2025-03-28GUANGZHOU HUAWEI TOMORROW SOFTWARE TECH
5 Cites 1 Cited by
Patent Information
- Application Number
- CN202411740109.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-29
- Publication Date
- 2025-03-28
- Estimated Expiration
- 2044-11-29
AI Technical Summary
Technical Problem
It is difficult for the prior art to accurately deal with similar or identical complaint subjects in the analysis of government complaint data, which affects the accuracy of statistical data.
Method used
The complaint analysis method based on LLM big model RAG technology is adopted, and the subject information of the complaint work order is extracted through the LLM big model, and similar subjects are recalled from the vector library using RAG technology, and the existence of the subject is determined through the LLM big model, and the database is updated to ensure data accuracy.
Benefits of technology
It effectively avoids misjudgment of similar objects or the same objects, improves the statistical accuracy of complaint data, and reduces the complexity of the algorithm.
✦ Generated by Eureka AI based on patent content.
Abstract
The invention discloses a complaint analysis method based on an LLM large model RAG technology. The complaint analysis method comprises the steps that S1, a new complaint work order is received; s2, extracting subject information of the complaint work order by using a large model; s3, recalling similar subjects from the subject information vector library by using an RAG technology; s4, judging whether the main body of the target complaint work order exists or not by using an LLM large model, if the main body of the target complaint work order exists, entering S5, and otherwise, entering S6; s5, updating a database work order table, and adding 1 to the number of associated subject complaint work order records corresponding to the subject of the target complaint work order; s6, inserting the main body of the target complaint work order into the database main body table as a new main body, adding 1 to the number of complaint work order records, and adding new main body information into the vector database; and S7, repeating the steps from S1 to S6, and processing all the work orders. According to the invention, misjudgment of similar objects or same objects can be prevented, and the accuracy of statistical data is high.
Need to check novelty before this filing date? Find Prior Art