Research On Tuple-Based Keyword Search In Relational Databases

With the developments of relational database technology and the maturation of keyword search technology over Internet, the problem of how to realize keyword search in relational databases becomes outstanding day by day, but because relational database system adopt structured storage mechanism, so data is stored in relations and tuples dividedly, this condition bring challenges to the realization of keyword search in relational databases. This paper analyses the theoretical basis and realization methods of relational database systems and keyword search technology, proposes content-based similarity calculate method and Minimum Steiner tree based AST query algorithm. The more details as follows.Firstly, introduce the concept, basis and major evaluation indexes of Information Retrieval, emphasize explain the principle and architecture of the Internet search engine. Introduce the principle of relation-based and virtual document based keyword search algorithm in relational databases, analyses and points out their advantages and disadvantages.Secondly, gives the concept of tuple graph and analyses the influence of the tuple structure in similarity calculation. Propose the content-based similarity calculate methods of attribute, tuple and edge. This make the tuple graph be a undirected and weighted graph which can be used to search immediately.Thirdly, proposes the concept of regional relevance in keyword search and regional storage in relational database system, discusses the influence of the matching between them in keyword search. Analyses the concept of Minimum Steiner tree, transfer keyword search into the evaluation problem of Minimum Steiner tree in tuple graph, propose Minimum Steiner tree based AST query algorithm, explain its basis and realization particularly. Finally, the experimentation is presented to testify the feasibility, validity of similarity calculate method and AST query algorithm. By compare recall and precision with existing tuple-based keyword search systems, demonstrate the advantage of the keyword search method which is proposed in this paper.