Research On Recognition And Classifying Method On Deep Web Query Interfaces

Recent years, a large amount of valuable data which hides inside on-line database behind Query Interfaces has been developed rapidly among Networks. We define this dynamic large amount of information which can't be searched through crawler a Deep Web. Therefore, how to provide customers with an Integrated Query Interface which face to the same field has become a hotspot during research field. Build Deep Web integration Interface include two questions .one is how to get Interface field what is affiliated with inquire ability means that identify Interface character information, another is classify by Interface field and inquire condition. To research at this two aspects, embody work as follows:At identify aspect, aiming at computing, maintenance and match matter, to bring forward a Interface identify arithmetic--SortIden base on label grouping. Different from tradition, at first, this way according to lable by abnormity character .Then, get every label group a independency unit. The method focuses on labels Subscript restriction, we can also say subscript adjacent as the clustering conditions. Bring forward the way of identifying simple attribute and complex attribute. Finally, through the two clustering effectively solves the problem of interface layer nested.Consider to diversity and complexity of Interface mode, when entire Interface is not in order, bring forward a Interface identify arithmetic—CuttingIden bases on two space. This way makes a Interface into different unit by range to label and difference of seeing. Support to a method that identify Interface character information by restricting abscissa and vertical.At classify aspect, to support a optimize classify implement—BayesOpt based on Bayes'classify, efficiency, nicety and calculate factor. To give design of classify node, bring forward the authority of node, array classify node by the authority. Giving a process of classify implement and self-study arithmetic of classify implement.At last, in order to check the effectiveness of the method, we have conducted a experiment to test all data. We concluded from the result that the method I have mentioned in this article has more advantages than any other methods. In addition, When we test effectiveness of classifying, we find that BayesOpt have more advantages in accurate rates and complexity of calculation.