论文标题
不同JSON表示对查询知识图的影响
The Effects of Different JSON Representations on Querying Knowledge Graphs
论文作者
论文摘要
知识图（kgs）已成为用于建模和查询数据集具有类似于图的结构的脱离事实标准。我们的重点是与查询KG相关的绩效挑战。我们为基于主题的名称/值（JSON-SNV），Triples（JSON-DT）和基于链条的名称/值（JSON-CNV）开发了三个基于信息等效的基于JSON的表示。我们通过将它们存储在两个基于文档的数据管理系统（DMS）上，即MongoDB和Couchbase并执行一组基准测试查询，从而分析了这些表示对查询性能的影响。我们还将执行时间与划分店Virtuoso，columner-columtor virtuoso和\ mbox {blazegraph}作为具有不同架构的三个主要DMS（aka，rdf商店）进行了比较。我们的结果表明，表示类型对查询执行有重大的性能影响。例如，JSON-SNV的表现要优于其他数量级，以执行主题对象加入查询。本文中提出的这一结果和其他结果可以帮助更准确地对新兴DMS进行基准测试。
Knowledge Graphs (KGs) have emerged as the de-facto standard for modeling and querying datasets with a graph-like structure in the Semantic Web domain. Our focus is on the performance challenges associated with querying KGs. We developed three informationally equivalent JSON-based representations for KGs, namely, Subject-based Name/Value (JSON-SNV), Documents of Triples (JSON-DT), and Chain-based Name/Value (JSON-CNV). We analyzed the effects of these representations on query performance by storing them on two prominent document-based Data Management Systems (DMSs), namely, MongoDB and Couchbase and executing a set of benchmark queries over them. We also compared the execution times with row-store Virtuoso, column-store Virtuoso, and \mbox{Blazegraph} as three major DMSs with different architectures (aka, RDF-stores). Our results indicate that the representation type has a significant performance impact on query execution. For instance, the JSON-SNV outperforms others by nearly one order of magnitude to execute subject-subject join queries. This and the other results presented in this paper can assist in more accurate benchmarking of the emerging DMSs.
