Research On Results Ranking And Clustering Of Keyword Retrival Over Relational Database

With the continuous development and extensive use of relational databasetechnology, the data stored in a relational database has become increasingly abundantand diverse. If user wants query information from those data quickly and conveniently,both keyword query in traditional information retrieval technology and structuredquery in relational database technology can not meet the user’s applicationrequirements. Therefore, realistic demand drives the development of relationaldatabase technology toward the direction of supporting more efficient query, such askeyword query.Currently, research on keyword search over relational database has become aresearch hotspot in the database fields, and has a broad application prospect. Since itallows ordinary users can simply enter keywords to get the related information theyneed, under the condition that not familiar with any query language (SQL statement)and not understanding the underlying knowledge model case. This paper studies andanalyses the current status of studies, working principle and implementation onkeyword search over relational databases. Based on this, this paper carries outresearch focusing on the search algorithms, sorting algorithms, and the resultsshowing. Aiming at solving the deficiencies in existing works, this paper proposes animproved query algorithm on map data based on Voronoi-path, a results rankingalgorithm based on the content and structure of keyword search over relationaldatabase, and a results clustering algorithm. The main work is as follows.Firstly, this paper introduces the composition of the system of keyword searchover relational database, and elaborates its working principle and summarizes theresearch status of each module, compares and analyzes their advantages anddisadvantages, and then determines the data model adopted in this research and givesintroduce on basic concepts.Secondly, research on query algorithm and ranking algorithm. Aiming at solvingthe deficiency of index redundancy in query algorithm on map data based onVoronoi-path, this paper proposes an improved algorithm. This algorithm uses intervalsearch, keyword filter and other methods to reduce the number of indexes. Theseimprovements not only can effectively reduce the number of indexes, but also canreduce the invalid results and retain the efficient results. As to the deficiencies of the scoring function in the existing methods, this paper proposes a new scoring functionto solve this problem.Thirdly, research on results clustering algorithm. In order to improve thedemonstrate effectiveness of the result interface, it is of great significance to clusterthe results before displaying to user. This paper proposes a new way of modeling thejoint tuple tree, and then presents a new results clustering algorithm, which can beapplied to all systems returning joint tuple tree as a result of relational databasekeyword retrieval.Finally, the experimental results show that the correctness and validity of theimproved algorithm based on Voronoi-path, and the results ranking algorithm, and theresults clustering algorithm. Comparison of the experiment results proves thesuperiority of the methods proposed in this paper.