A complaint classification method based on natural language

By building an emotion classifier model in the complaint management system and using the GNN network structure to classify complaint content, and obtaining user complaint records in combination with the RAG knowledge base, the problem of relatively general performance of AI natural language models in the existing technology is solved, and the efficient, accurate and universal effects of complaint handling are achieved.

CN119166814BActive Publication Date: 2025-03-25STONE TECH CO LTD
2 Cites 0 Cited by

Patent Information

Application Number
CN202411255303.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-09
Publication Date
2025-03-25
Estimated Expiration
2044-09-09

AI Technical Summary

Technical Problem

In the complaint management system, the customized development of AI natural language models has led to relatively average performance. Users are disgusted with intelligent robot dialogue, and still require manual intervention, and are less versatile.

Method used

The complaint classification and discrimination method based on natural language is adopted, and the emotional classifier model is constructed, and the historical complaint work ticket text content is used to perform word segmentation, word embedding processing and emotional label training, and user complaint records are obtained in combination with the RAG knowledge base to achieve accurate classification of complaint content.

Benefits of technology

It improves the efficiency and accuracy of complaint handling, optimizes the user's complaint sentiment analysis capabilities, can quickly locate and identify difficult questions and escalate complaints, improves the efficiency of complaint identification, and has strong versatility.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present invention relates to the field of data processing technology, and in particular to a complaint classification and discrimination method based on natural language; the method comprises the following steps: S1, constructing a sentiment classifier model; S2, obtaining the text content of a complaint work order to be processed, obtaining a corresponding plurality of word embedding vectors, and obtaining the sentiment label of the word corresponding to each word embedding vector; S3, obtaining the previous personal complaint records of the user and the complaint records of others of the same type as the previous personal complaint records; S4, integrating the complaint content, complaint type, personal complaint record and complaint records of others for storage; S5, converting the stored content into text form; S6, marking the sentiment label for each word in the complaint content converted into text form, and completing the complaint classification; the present invention classifies the vocabulary in the complaint content more accurately, effectively improves the complaint recognition efficiency, and has strong versatility.
Need to check novelty before this filing date? Find Prior Art