| In recent years, the Internet is booming in the world and the information it carriesgrowth much faster. The information carried by the Internet can be divided into twoparts, one part called Surface Web is accessible directly on the web and the other calledDeep Web is stored in the database behind the Web in a structured way which makesDeep Web data more valuable. Deep Web can be accessed only through the queryinterface provided by the Web, however, increasing number of web applications need toaccess query interfaces frequently and automatically. Therefore, it is necessary formodeling the semantics and structure of the query interface.Concept lattice theory, also known as Formal Concept Analysis, was proposed as amathematical theory by Rudolf Wille in1982. Concept lattice theory is mainly used forthe formalization of knowledge, but it is also an effective tool for data analysis. The keyissue of query interface modeling is to discover semantic relations, furthermore, the waywhich concept lattice represent and organize knowledge has been proved to be verysuitable for mining semantic relations, so the concept lattice theory is an effective toolto recognize semantic relations.In order to complete the query interface modeling, a Deep Web query interfacemodel which clearly describes the organization structure and semantic relationshipsbetween elements of the query interface was proposed in this paper. For this model, thepaper mainly focuses on the methods of query interface modeling. First, we defined alayout matrix of query interface and related algorithms were given. Second, thedimensional spaces that were used to analyze query interface were proposed,furthermore, its correctness and validity had been proved. Third, the paper studied howto use concept lattice theory to discover semantic relations in the query interface.Finally, based on the model above, we implemented a Deep Web query interfacemodeling system. The system consists of HTML parsing module, layout matrixoperation module, dimensional space mapping module and semantic rules miningmodule. Through the analysis of execution time, modeling precision and recall, weverify that the proposed modeling system is valid. |
