利用关系数据库系统对半结构化数据进行近似查询

提出一种利用关系数据库系统在一般图结构的半结构化数据上进行近似查询的途径.根据嵌套结构和文本值的相似性来度量路径的相似性;根据路径的相似性得到查询目标节点与数据源节点的相似性.为返回数据源中与查询目标节点相似的节点,首先提取出数据源中长度在固定范围内的所有路径,然后利用关系数据库系统将其与查询路径进行相似性连接,并按相似度从大到小返回所有结果.为提高相似性连接的效率,引入q窗口概念,并利用若干路径相似的必要条件来减少计算相似性函数的次数.试验证明了其有效性.

Approximate Querying of Semistructured Data Using a Relational Database System

An approach to approximate querying of general graph structured semistructured data is proposed based on a relational database system. A similarity measure for paths based on nesting structures and text values was brought out, from which the similarity between the target query node and a data source node was derived. To get the source nodes similar to the target query node, firstly the paths whose lengths were within an interval were extracted from the data source, then a similarity join process between them and the query paths was carried out using a relational database system. Finally the query result nodes were returned in a descend order of their similarity to the target query node. To make the similarity join process more efficient, the concept of q-windows was introduced and several necessary conditions were used to decrease the time of calculating costly similarity function. The experiments prove the effectiveness of the approach.