基于网络论坛数据的未满足用户需求挖掘方法研究  

Unmet User Needs Mining Method Based on Web Forum Data

作　　者：李奕潼 徐照光 党延忠[1] LI Yitong;XU Zhaoguang;DANG Yanzhong(Dalian University of Technology,Dalian,Liaoning,China)

机构地区：[1]大连理工大学经济管理学院,辽宁省大连市116024

出　　处：《管理学报》2025年第1期125-134,共10页Chinese Journal of Management

基　　金：国家自然科学基金资助项目(72001034);辽宁省博士科研启动基金计划资助项目(2022-BS-088);中央高校基本科研业务费引进人才科研专题资助项目(DUT23RC(3)037)。

摘　　要：为帮助企业改进产品,提出一种基于网络论坛数据的两阶段用户需求挖掘方法。第一阶段利用卷积神经网络模型,识别包含未满足用户需求的文本;第二阶段采用双向长短期记忆网络-条件随机场模型,从未满足用户需求的文本中提取用户观点四元组(主题,对象,属性,属性值),并转化为用户需求四元组(主题,对象,属性,属性期望值)。选取汽车之家论坛上某车系数据进行实验,研究表明,该挖掘方法通过构建四元组结构来表示用户需求,从细粒度层面展示产品特征,并逐层钻取需求细节,减小需求不确定性,明确具体的用户需求,从而提高产品竞争力。并通过比较说明该方法的可行性和有效性。To assist enterprises in improving their products,a two-stage user demand mining method based on online forum data is proposed.In the first stage,a convolutional neural network model is used to identify texts that contain unmet user demands.In the second stage,a Bi-directional Long Short-Term Memory Network-Conditional Random Field(BiLSTM-CRF)model is employed to extract user opinion quads(topic,object,attribute,attribute value)from the texts with unmet user demands and transform them into user demand quads(topic,object,attribute,expected attribute value).Experiments were conducted using data from a specific car series on the Autohome forum.The study shows that the mining method constructs a quad structure to represent user demands,demonstrates product features at a fine-grained level,and drills down into the details of the demand layer by layer to reduce the uncertainty of the demand and clarify the specific user needs,thereby enhancing the product’s competitiveness.The feasibility and effectiveness of the method are also demonstrated through comparisons with other methods.

关 键 词：用户需求 网络论坛 深度学习 用户生成内容 

分 类 号：C93[经济管理—管理学]