Being different from general retrieve methods which load data graph into main memory one time, a new method of keyword search over relational database based on graph partition on data graph is proposed in the paper. A large data graph can be partitioned to a set of subgraphs which have simple connectivity and are stored in cache. Then related subgraphs are obtained through supergraph search algorithm. Only that related subgraphs are loaded into main memory and recovered to initial structure while searching for candidate results based on keywords. Then final results are obtained by ergodic subgraphs. The results conducted from the experiments show that the naive approach can address problems of processing large graph effectively when it can not be loaded into main memory one time and enable keyword search over relational database applied to large and complicate data graph without main memory constraint.%不同于常用的一次性将图加载入内存的检索方法,提出了一种基于数据图划分的关系数据库关键词检索方法,该方法首先将大图细分成多个相互之间具有简单连通性的小图并存储于缓存中,然后通过超图查询算法获取与查询条件相关的子图,进行关键词检索获取候选结果时,只需加载相关子图并恢复子图的初始结构,通过遍历子图获得最终结果.实验表明,该方法可以有效解决对无法放入内存的大型数据图的处理问题,使得关系数据库关键词检索实用于庞大而复杂的数据图中而不受内存约束.
展开▼
