Abstract: The paper presented an algorithm which fills the ctuery interface by using machine learning based on the analysis of mechanism of Deep Web query. The algorithm is able to extract data automatically. Firstly, a 2D table is constructed. The columns of the table are controllers extracted from pages of the Deep Web query interface. Then values of the table are filled by giving values to all the controllers. Next, a learning of classification is going to be achieved according to the result whether the extraction of data successfully or not. Finally, the data is extracted by constructing request string automatically through the results of the learning. The experiment shows that the algorithm runs effective1y.

Key words: Data extraction,Semantic,Machine learning,Deep Web