CLASCN: Candidate Network Selection for Efficient Top-k Keyword Queries over Databases

Keyword Search Over Relational Databases (KSORD) enables casual or Webusers easily access databases through free-form keyword queries.Improving the performance of KSORD systems is a critical issue in thisarea. In this paper, a new approach CLASCN (Classification, Learning AndSelection of Candidate Network) is developed to efficiently performtop-k keyword queries in schema-graph-based online KSORD systems. Inthis approach, the Candidate Networks (CNs) from trained keywordqueries or executed user queries are classified and stored in thedatabases, and top-k results from the CNs are learned forconstructing CN Language Models (CNLMs). The CNLMs are used tocompute the similarity scores between a new user query and the CNsfrom the query. The CNs with relatively large similarity score,which are the most promising ones to produce top-k results, will beselected and performed. Currently, CLASCN is only applicable forpast queries and New All-keyword-Used (NAU) queries which arefrequently submitted queries. Extensive experiments also show theefficiency and effectiveness of our CLASCN approach.