基于pay-as-you-go模式的海量RDF图数据的关键词查询

With the emergence of massive RDF data, query over massive RDF data has become the important issues in the field of Big Data. However, the existing research has ignored the ambiguity of keyword query, and the support for dynamic data query. To address above problems, this research project will focus on keyword query processing and optimization over massive RDF graph data in a pay-as-you-go fashion. First of all, the dynamic evolution algorithm of RDF graphs is presented by analyzing the characteristics of RDF data evolution, and then the evolutionary computation algorithm is proposed, after which, the appropriate data constraints are defined to ensure the accuracy and consistency of data description in the dynamic evolution of RDF graphs. Secondly, under the guidance of dynamic evolution mechanism, an approach based on pay-as-you-go fashion to annotate semantic description in RDF graphs is also proposed. Through data mining and user feedback, this approach determines the key information (e.g., key nodes and key edges) which affect the query result, and then the distributed parallel computing environment is constructed to extract the changes of key information to enhance RDF semantics incrementally. Thirdly, RDF's query algebra based on the query theory of graph structure model is studied, which supports the association query based on the edge (data association), and then the distributed delay calculation method is designed to realize the pay-as-you-go keyword query. Finally, the validity of the work in this project is examined by the evaluation mechanism which combines investment income theory and information retrieval evaluation method.

随着海量RDF数据的涌现，如何准确、快速的查询RDF数据成为大数据研究领域急需解决的重要课题。针对现有研究存在的缺乏支持动态数据查询、查询语义歧义性大的难题，本项目拟研究面向海量RDF图数据的pay-as-you-go(PAYG)模式关键词查询。首先从机理上探究RDF图动态演化过程，研究演化计算方法，定义合适的数据约束保证RDF图动态演化时数据描述的准确性和一致性。其次，以该动态演化机理为指导，设计PAYG模式语义动态添加方法：通过数据挖掘、用户反馈等方式判断影响查询效果的关键信息(如关键节点、关键边)，通过增量提取关键信息的变化内容来渐进增强RDF语义的描述。同时，借鉴图结构模型的查询理论构建RDF图查询代数，支持以边(数据关联)为基础的关联查询，设计分布式延迟计算方法实现PAYG模式关键词查询。最后融合投资收益理论和信息检索评价方法，构建PAYG模式关键词查询的评估机制。

随着海量RDF数据的涌现，如何准确、快速的查询RDF数据成为大数据研究领域急需解决的重要课题。针对现有研究存在的缺乏支持动态数据查询、查询语义歧义性大的难题，本项目研究了面向海量RDF图数据的pay-as-you-go(PAYG)模式关键词查询，顺利完成原定计划，取得了预期的效果，具体工作如下：1）研究了RDF图动态演化过程及其演化计算方法。2）设计了PAYG模式语义动态添加方法，通过增量提取关键信息的变化内容来渐进增强RDF语义的描述。3）构建RDF图查询代数，支持以边(数据关联)为基础的关联查询，设计分布式延迟计算方法实现PAYG模式关键词查询。4）构建了PAYG模式关键词查询的评估机制，对上述工作进行改进和优化。本项目的研究能够提高海量RDF图的查询性能，对图模型的查询研究有一定的借鉴和参考价值。