文摘
Deep web is the web that is dynamically generated from data sources such as databases or file systems. Crawling deep web is the process of collecting hidden data by issuing appropriate queries in order to download most of the data. Our main challenge is to select appropriate queries in order to obtain most of the data from a data source. A naive solution, which selects the queries that return most results, is problematic because (1) the results may not cover the data source, and more importantly, (2) the results suffer from high overlap, which makes the acquisition of new data items almost impossible after certain steps. The thesis experiments with four different algorithms to select the queries that minimize the overlap rate: (1) greedy algorithm based on set packing; (2)cluster-based algorithm to remove the queries that result in similar returns.;Keywords: deep web, hidden web data discovery, data mining, clustering, information retrieval.