There is a significant interest in the areas of data mining about the need to improve existing techniques for extracting exact data from deep web, disconnecting the services, to prevent and reduce the response time to user. In this project, a new framework as interface between user request and database in deep web is designed. User request is sent by a web form to search for information such as books, cars and so on. Many websites such as Amazon, has such forms. Requests sent by users to deep web data bases, in this way request reach to framework. Then framework sends the data request to databases according to user requests. Then, framework gets one dataset that may include records that match with user request. After three level process, framework extracts the exact records according to user request and send to the user interface. The project focuses on challenges of data recovery from web databases. Experimental results in this project are two types. First experimental results extract a standard and exact data from databases in depth of web. Second experimental results are estimating the number of process level of discovery algorithm in eight domain of the QI whole the Web