摘要
【目的/意义】基于专业性多源网络数据,构建同质性或异质性领域知识图谱。【方法/过程】基于场景性关键词相似度计算进行同质网络层次聚类,揭示症状的同质性网络层次以及治疗方案的同质性网络层次;基于2-mode共现矩阵进行异质网络聚类,兼顾症状及治疗方案的关联关系,通过设置不同阈值揭示其异质关联性网络层次。【结果/结论】基于聚类纯度及熵值评价指标,实验结果显示:就纯度评价指标而言,进行同质性网络知识图谱分析较为合理;就熵值指标而言,进行异质性网络知识图谱分析较为合理。
【Purpose/significance】Based on professional multi-source network data to construct homogeneous or heterogeneous networks domain knowledge graph.【Method/process】Hierarchical clustering of homogeneous networks based on similarity computation of scene keywords was conduct, the homogeneity network of symptom and treatment plan were revealed; Heterogeneous network clustering based on 2-mode co-occurrence matrix was condect, association between symptoms and treatment options was considered, setting different thresholds to reveal the hierarchy of heterogeneous networks.【Result/conclusion】Based on the clustering purity and entropy evaluation index, the experimental results show that： in terms of the purity evaluation index, the construction of homogeneous network knowledge map is more reasonable; in terms of entropy index, it is reasonable to build heterogeneous network knowledge graph.
作者
李保珍
苏菁
LI Bao-zhen1, SU Jing2(1.Government Audit Big Data Research Centre, Nanjing Audit University, Nanjing 211815, China; 2.School of Economics and Management, Jiangsu University of Science and Technology, Zhenjiang 212003, China)
出处
《情报科学》
CSSCI
北大核心
2018年第10期13-19,共7页
Information Science
基金
国家自然科学基金项目(71673122
71273121)
关键词
知识图谱
同质网络
异质网络
聚类
knowledge graph
homogeneous network
heterogeneous network
clustering
