融合病情特征的在线健康社区用户需求识别模型研究  

Research on User Needs Identification Model of Online Health Communities Integrating Disease Characteristics

作　　者：成全 林颖茹 Cheng Quan;Lin Yingru(School of Economics and Management,Fuzhou University,Fujian Fuzhou 350116)

机构地区：[1]福州大学经济与管理学院,福建福州350116

出　　处：《情报理论与实践》2025年第4期125-134,共10页Information Studies:Theory & Application

基　　金：国家社会科学基金项目“多源在线医疗健康信息的语义融合及精准推荐服务研究”的成果,项目编号:19BTQ072。

摘　　要：[目的/意义]针对在线健康社区用户生成内容,提出了融合病情特征的多任务用户需求识别模型(MUNI-DC),深入挖掘用户需求,形成由提问意图和提问实体两部分内容组成的用户需求主题体系。[方法/过程]通过构建BERT-wwm模型对用户需求数据进行预训练,融合病情特征实现对用户意图的识别,采用多层标签指针网络实现对在线健康社区用户提问数据的实体识别,以此为基础完成对在线健康社区用户需求的识别任务。[结果/结论]对比实验发现,相比于单一任务模型,本模型在用户需求识别结果的precision、recall、F1等指标上均有所提升;消融实验发现,融合病情特征和多层标签指针网络能够有效提升模型的用户需求识别效果,所提出的融合病情特征的多任务用户需求识别模型在应对在线健康社区用户信息需求分析任务中具有参考价值。[Purpose/significance]A multi-task user needs identification model integrating disease characteristics(MUNI-DC)is proposed for user generated content in online health communities.Deeply explore user needs and form a demand theme system consisting of two parts:questioning intention and questioning entity.[Method/process]By constructing a BERT-wwm model for pre-training user demand data and integrating user medical condition characteristics to achieve recognition of user intentions;Furthermore,a multi-layer label pointer network is used to achieve entity recognition of user query data in online health communities.Based on this,identify the needs of users in online health communities.[Result/conclusion]Comparative experiments have shown that compared to a single task model,this model has improved in precision,recall,F1,and other indicators of user requirement recognition results.The ablation experiment found that the fusion of disease characteristics and multi-layer label pointer network can effectively improve the user demand recognition performance of the model.The proposed MUNI-DC model has reference value in dealing with online health community user information demand analysis tasks.

关 键 词：在线健康社区 用户需求识别 实体识别 语义识别 知识图谱嵌入 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]