基于中文电子病历知识图谱的实体对齐研究

Entity Alignment Based on Knowledge Graph of Chinese Electronic Medical Record

医疗知识图谱中知识重叠和互补的现象普遍存在,利用实体对齐进行医疗知识图谱融合成为迫切需要。然而据作者调研,目前医疗领域中的实体对齐尚没有一个完整的处理方案。因此该文提出了一个规范的基于中文电子病历的医疗知识图谱实体对齐流程,为医疗领域的实体对齐提供了一种可行的方案。同时针对基于中文电子病历医疗知识图谱之间结构异构性的特点,该文设计了一个双视角并行图神经网络(DuPNet)模型用于解决医疗领域实体对齐,并取得较好的效果。

Entity alignment is essential to fuse the medical knowledge graphs since the phenomenon of knowledge overlap and complementarity is common in different medical knowledge graphs.However,according to our research,there is not yet a complete solution for entity alignment in the medical field.Therefore,we propose a standardized entity alignment process based on the Chinese electronic medical record knowledge graph,which provides a feasible scheme for entity alignment in the medical field.Meanwhile,according to the characteristic of the structural heterogeneity of the medical knowledge graph,we design a Dual-view Parallel Graph Neural Network(DuPNet)to solve the problem of entity alignment in the medical field,which achieves good results.