A Hybrid Method for Extracting Deep Web Information

Some previous works show that more than 60% of the information available on the Web is located in Deep Web database. Such information cannot be directly indexed by search engines. In this paper, a hybrid method, which is composed of a domain model and a block importance model is proposed to extract information in Deep Web. The domain model is used for classifying and identifying whether a form is a WQI. The block importance model is used for filtering noisy information in response pages. These two models are both compared with a rule-based method. The experiment results indicate that the domain model yields a precision6.44% higher than that of the rulebased method, whereas the block importance model yields an F1 measure 10.5% higher thanthat of the XPath method.