基于大模型的电商平台用户评论需求获取与情感分析研究

Research on Demand Extraction and Sentiment Analysis from User Reviews on E-Commerce Platforms Based on Large Language Models

摘要: 在互联网时代，电商平台用户评论数据呈快速增长态势，如何从海量文本中挖掘用户需求并分析不同需求下隐含的用户情感，成为企业提升产品竞争力的关键。本研究提出一种基于大语言模型(LLM)的链式分析框架，设置三个代理模块系统性地实现用户需求获取与情感分析。三个代理模块分别完成从海量非结构化用户评论中精准提取核心需求陈述、基于提取的需求文本进行细粒度情感倾向分析、需求的多维度分类这三个任务。实验表明，利用大语言模型实现链式分析，不仅克服了单模型处理复杂任务的局限性，还通过代理间的信息流传递形成闭环分析链路。本框架为企业和研究者提供了可扩展、自动化的在线评论的深度分析工具，对精准把握用户诉求、优化产品设计及制定营销策略具有一定的实践意义。

Abstract: In the internet era, user review data on e-commerce platforms has experienced explosive growth. How to mine user demands from massive texts and analyze the implicit user emotions under different demands has become a key factor for enterprises to enhance product competitiveness. This study proposes a multi-agent collaboration framework based on Large Language Models (LLMs), incorporating three agent modules to systematically achieve user demand extraction and sentiment analysis. These three agent modules respectively accomplish three tasks: accurately extracting core demand statements from massive unstructured user reviews, conducting fine-grained sentiment tendency analysis based on the extracted demand texts, and performing multi-dimensional classification of demands. Experiments show that the agent collaboration mechanism implemented by large language models not only overcomes the limitations of single models in handling complex tasks but also forms a closed-loop analysis chain through information flow transmission between agents. This framework provides enterprises and researchers with an extensible and automated tool for in-depth UGC (User-Generated Content) analysis, and holds significant practical implications for accurately grasping user needs, optimizing product design, and formulating marketing strategies.