不确定RDF知识图谱数据查询关键技术研究

知识图谱是人工智能的核心基础设施，已将RDF数据模型作为其数据层事实上的标准。与此同时，由于噪声等不确定因素存在，使得知识表达难以完备化，RDF知识图谱中出现了不确定信息。传统的RDF数据模型与查询方式不能满足不确定RDF知识图谱的应用需求，扩展RDF模型来管理知识图谱数据中的不确定信息成为一项紧迫又富有挑战的任务。为此，申请项目以概率论作为理论基础，使用可能世界模型描述RDF知识图谱数据，构建不确定RDF知识图谱数据模型。基于该模型，研究不确定RDF知识图谱上基于正则表达式的导航式路径查询、基于路径索引的子图模式匹配查询和基于r-clique的关键字分析型查询等3类基本的图查询处理方法。通过对模型定义、索引构建、图查询处理、剪枝技术和抽样算法等关键技术的研究，为不确定RDF知识图谱数据建模和查询处理提供理论方法和技术，为智能环境下开展知识图谱应用研究奠定坚实的基础。

Knowledge graphs are the core infrastructure of artificial intelligence, and the RDF data model is the de facto standard for the data layer. At the same time, due to the existence of uncertain factors such as noise, it is difficult to complete the expression of knowledge, and additional uncertain information appears in the RDF knowledge graph data. The traditional RDF data model and query method cannot meet the application requirements of uncertain RDF knowledge graph. Establishing a standard model to manage uncertain information in RDF knowledge graph data has become an urgent and challenging task. Therefore, the proposed project uses probability theory as the theoretical basis, uses the possible world model to describe the RDF knowledge graph data, and constructs an uncertain RDF knowledge graph data model. Based on this model, three major types of graph query processing issues, such as regular expression-based navigation path query, path index-based subgraph pattern matching query, and r-clique-based keyword analysis query, are studied on the uncertain RDF knowledge graph. Some key techniques such as model definition, index construction, graph query processing, pruning technology and sampling algorithm will be studied, which can propose theories and techniques for data modeling and query processing over uncertain RDF knowledge graph data, and a solid foundation can be put for the application of knowledge graph in the intelligent environment.