基于多标签节点和LINE的知识图谱动态更新方法

Dynamic Updating Method of Knowledge Graph Based on Multi-Label Nodes and Line

针对现有知识图谱动态更新方法只考虑节点的单类标签而忽略来自外部的丰富标签信息问题,本文定义了以命名规则为基础的标签结构模型,利用大规模网络嵌入方法(large-scale information network embedding, LINE)预测节点间的关系链接,提出了以多标签节点为主的动态多元属性标签方法(dynamic multivariate attribute labeling, DWAL)。为了验证该方法的可行性和有效性,对知识图谱三元组进行统计和校验,实现知识图谱质量控制,并在新技术需求数据集上进行测试,最后采用Neo4j软件更新知识图谱,在多数据集上验证该方法在重复节点处理问题上的有效性。实验结果表明,该方法能够为多标签节点合理的增加标签信息,扩大节点信息量,并有效地去除冗余,减少重复节点的构建,为知识图谱下一步研究打好坚实基础。该研究对提高知识图谱动态更新的准确性具有重要意义。

In order to solve the problem that the existing dynamic updating methods of knowledge graph only consider the single class label of nodes and ignore the multi-label information from the outside, the label structure model is defined based on multi-label nodes and DWAL(the dynamic multivariable attribute labeling method) is proposed, using LINE method to predict the relationship link between nodes in knowledge graph. In order to verify the feasibility and the effectiveness of the method, neo4 j software is used to update the knowledge graph. The effectiveness of the method in repeated nodes is verified though NTAR data set and the feasibility is verified on the other data sets. Experimental results show that the method can reasonably increase label information for multi-label nodes, effectively remove redundancy and reduce the construction of duplicate nodes. This research has a great significance to improve the accuracy of dynamic updating of knowledge graph.