Deep Web (often called hidden web or invisible web) is composed of all the web databases. With the evolution of the "deep web", more and more researchers pay attention to the "integration" of the web database. However, to achieve this goal, it needs a complex system and many applications to work together. We are interested in an automatic extracting system to get the formulas or the lists of the results from those websites in the specific domain of government procurement. To tackle this challenge, we propose a solution to create a unified interface and to inquire resources in a predefined domain. In this paper, we will discuss the automatic extracting system in several steps. First of all, the web query interfaces crawler which can execute JavaScript guarantees the coverage of the web database. Secondly, the query interface extractor and the interface integrator can allow us to query all these founded web databases through a global query interface. Thirdly, the result page extractor and the result integrator can give a unified presentation. Lastly, a feedback method is developed to gather the result accuracy. A statistical model is built to improve the performance of steps 2 and 3. We assume our system is a dynamic system, which means the more we use it, the better results we will get.