Telecom Marketing Recommendation Script Generation Based on LLM Structure

This paper presents a dialogue generation technique based on large model retrieval enhancement, which fully utilizes users' unique information and marketing knowledge to generate optimal marketing language. The technical proposal includes speech recognition, emotion recognition, user behavior analysis, knowledge retrieval, prompt construction, large model processing, user feedback collection, and performance evaluation, which forming a complete system process. The key technical points include language generation technology based on large models and retrieval enhancement, the combination of natural language processing and large-scale text data, the construction of industry-specific knowledge bases, unique Prompt construction methods, and LoRA fine-tuning technology. These technical points together ensure the personalization and accuracy of marketing speech, with an accuracy rate increased from 53.77% to 87.95%. Through a comprehensive review and analysis of existing technologies, it is evident that this article possesses distinct advantages over the closest prior art in terms of integrating user information and industry knowledge. It effectively leverages users' exclusive infor-mation and discourse knowledge base to craft more professional, targeted, and optimal marketing discourse, ultimately enhancing user satisfaction and driving business growth.