基于MuGNN模型的互联网医疗知识融合研究  

Research on Internet Medical Knowledge Fusion Based on MuGNN Model

作　　者：廖开际[1] 王莹 LIAO Kaiji;WANG Ying(School of Business Administration,South China University of Technology,Guangzhou 510641,China)

机构地区：[1]华南理工大学工商管理学院,广州510641

出　　处：《河南科学》2021年第12期2014-2022,共9页Henan Science

摘　　要：为了解决多来源医疗知识库融合过程中常见的知识冗余问题,基于综合多种注意力机制和图卷积神经网络的MuGNN模型对互联网医疗知识融合的效果进行了研究.以乳腺癌疾病为例,首先构建了基于不同医疗网站的疾病实体关系库,然后利用MuGNN模型完成了实体对齐,同时与JAPE模型和GCN-Align模型的实体对齐效果进行了对比,最后对基于不同医疗网站的疾病实体关系库进行知识融合并通过Neo4j图数据库对融合后的知识图谱进行可视化处理.结果表明,与JAPE模型和GCN-Align模型相比,MuGNN模型的实体对齐效果更好.利用综合多种注意力机制和图卷积神经网络的MuGNN模型对互联网医疗知识进行融合,有助于提升多来源互联网医疗知识的融合效果,有助于多源知识图谱的构建与补全,有助于提供更优质的知识服务.In order to solve the problem of knowledge redundancy in the process of multi-source medical knowledge database integration,the effect of internet medical knowledge fusion based on Mu GNN model with multiple attention mechanisms and graph convolutional neural network was studied. Taking breast cancer as an example,a disease entity relationship library based on different medical websites was constructed first,and then the entity alignment was achieved by using Mu GNN model. At the same time,the entity alignment effect of Mu GNN model was compared with that of JAPE model and GCN-Align model. Finally,the knowledge fusion of disease entity relationship databases based on different medical websites was carried out,and the fused knowledge map was visualized through Neo4 j graph database. The results show that compared with JAPE model and GCN-Align model,MuGNN model has better entity alignment effect. The Mu GNN model integrating multiple attention mechanisms and graph convolutional neural network is used to fuse internet medical knowledge,which is helpful to improve the fusion effect of multi-source internet medical knowledge,to construct and complete the multi-source knowledge map,and to provide better knowledge services.

关 键 词：互联网医疗 知识融合 实体对齐 注意力机制 图卷积神经网络 

分 类 号：G203[文化科学—传播学] TP391[自动化与计算机技术—计算机应用技术]