基于图划分的领域本体RDF存储方法

A domain ontology RDF storage method based on graph partitioning

针对海量RDF图数据分布式存储无法有效保持数据语义结构完整性的问题,提出一种基于标签传播和标签能量函数的多级图划分方法。该方法首先对领域本体解析所得的RDF图进行顶点ID标识,并为实例数据的主语分配初始标签;利用标签传播方法对各顶点进行标签设置形成语义结构相似的顶点集合;在此基础上通过多级图粗化和标签能量函数限制顶点集合的大小实现数据的语义分区;将该方法应用于民航突发事件领域本体的分布式存储与查询。采用边割率对领域本体数据的分区效果进行了分析与比较,实验表明该方法在减少边割率的基础上,保证了查全率并同时提高了民航突发事件相似案例的查询效率,为大规模领域本体的分布式存储与语义查询提供了进一步的方法支持。

As the distributed storage of massive RDF graph data cannot effectively maintain the semantic structure integrity of data,a multi-level graph partitioning method based on the label propagation and label energy function is proposed.In the method,the ID identification of vertexes is conducted for the RDF graph obtained by parsing the domain ontology.The initial la-bel is assigned to the subject of the instance data.The label propagation method is used to set the label for each vertex,so as to form a vertex set with the similar semantic structure.On this basis,the size of the vertex set is limited by means of multi-level graph coarsening and the label energy function,so as to realize semantic partitioning of data.The method is applied to the dis-tributed storage and query for the emergency domain ontology of civil aviation.The edge cut rate is used to analyze and compare the partitioning effect of domain ontology data.The experimental results show that the method can guarantee a high recall rate on the basis of reducing the edge cut rate,and improve the query efficiency of similar emergency cases in civil aviation,which can provide a further methodology support for distributed storage and semantic query of large-scale domain ontology.