Research On Fuzzy Keywords Search Over Relational Databases And Its Optimization

Traditional relational databases query must be given the name of attribute, which can not be the same as a search engine keyword research freely. Keywords search over relational databases (KSORD) can query in relational database freely, without knowledge of relational databases structure. After analyzing and researching KSORD, this paper proposed a solution for two the deficiency of current KSORD.Because the current KSORD has only accuracy query, this paper proposed a solution of vague keywords search over relational databases. This solution gives a high frequency word generating algorithm, which finds high frequency words from table of database. The words could be stored in the table called "hotkeys". Keywords correcting algorithm is gave, through which, the corrective keywords can be found from table "hotkeys". After the corrective keywords are submitted to KSORD, the ideal results are found.In addition, because current KSORD is low efficiency, so this paper proposed a solution based on the table of join tree. In schema graph, the PageRank value is be added. Then a formula is given, through which the weight is added in edge. Giving a sub-graph which of node is less than Maxsize generating algorithm and a minimum spanning tree and "select" template generating algorithm, through which the join trees which node are less than Maxsize can be found in schema graph. These join trees could be stored in a table called "joinTree". During the query, the candidate join trees are found from the table "joinTree" instead of the schema graph by BFS.Finally, two KSORD experiments has been designed in this paper, which is used to verify a solution of vague keywords search over relational databases and a solution based on the table of join tree. By the recall, precision and query time as an indicator, a solution of vague keywords search is verified feasibility and effective.Then compare with a solution based on the table of join tree and a solution of BFS algorithm. The query time can be shortened, while the recall and precision don’t reduce. With the scale of relational database increasing, the optimized effect is more obvious.