Research on Automate Discovery of Deep Web Interfaces

The main means to obtain information from Deep Web is submitting query condition through the provided query interfaces, so it is the first problem that needs to be solved for Deep Web data integration system. At present, most researchers think of query interface is merely defined within the form html tag. This paper firstly proposes the concept of interface block, then designs the interface block location method based on page and vision information, and finally takes the judgment of whether interface block is a query interface or not as the special multi-class classification problems and by applying classification algorithm combining C4.5 decision tree and SVM. The experiment adopts TEL-8 data sets of UIUC, and the findings indicate that the method in this paper get an accuracy of 97.30%, and has good feasibility and practicability.