大模型驱动下的网络投诉自动化分类与智能响应系统设计

打开文本图片集

中图分类号：TN915.07；TP311 文献标识码：A 文章编号：2096-4706（2025）19-0121-05

Abstract： With the expansion of user scale and the increasing service complexity in the telecommunications industry， traditional networkcomplaint handling models face challenges suchas ineffciencyand highcosts.This paper proposes a large model-driven automated clasification and intelligent responsesystem for network complaints.By leveraging the Model Context Protocol（MCP），thesystemachieves multimodaldata fusionanddynamictask orchestration，andconstructsafourlayercollaborativearchitecturecomprising prception，analysis，decision-making，andexecutionlayers.Thesystemitegates OCR，ASR，andNLPtechnologies for multimodalcomplaintanalysis，combines Retrieval-Augmented Generation （RAG）with reinforcementlearningalgorithms tooptimizedecision-makingprocesses，andsupports high-concurrencyprocessngofupto100 000 daily work orders.Empirical results demonstrate thatthe systemreduces response time tolessthanone minute，improves classification accuracy to over 85% ，and cuts labor costs by 50% ，providing a viable pathway for intelligent transformation of the telecommunications industry.

Keywords： large model; multimodal fusion; Model Context Protocol; Retrieval-Augmented Generation

0 引言

当前通信行业正处于数字化转型的关键阶段，用户对服务响应速度和问题解决效率的要求日益提高。（剩余6321字）