Multiobjective Optimization Models And Solution Algorithms For Query Interfaces Integration

Deep Web is a set of resources that hide in the Web database and can not beaccessed through traditional query interfaces in the networks. For a domain-specificdeep web, different network stations provide different query interfaces with differentquerying capabilities. Through integrating domain-specific query interfaces, almost allresources in this domain can be accessed. The integration of query interfaces containsschema matching and schema integration, and schema matching has been studiedwidely. In this paper, the schema integration of query interface integration problem wasaddressed.The differences among the real query interfaces bring many difficulties to obtain asuitable interface which can satisfy all requirements when integrating the queryinterfaces. In order to meet most requirements, the frequent structure of queryinterfaces is used to construct integrated query interface. Aiming to obtain awell-structured and semantic comprehensive integrated query interface, two novelmulti-objective optimization models are proposed for the schema integration in thequery interface integration problem, and a solution algorithm is also designed for eachmodel, respectively.In the first model, an ordered tree was employed to represent a query interface,which can describe the structural information of the query interface. Based on theordered tree, a new kind of structural constraints and ordering constraints are defined.An integrated query interface should integrate the structure used frequently in all queryinterfaces. Thus the query interface integration problem can be converted to theproblem of searching frequent structure from all query interfaces, that is, searching afrequent subtree from all ordered trees of query interfaces. At last, a novelmulti-objective optimization model with maximizing the number of leaf nodes and thefrequency of the frequent subtrees was built, and the related solution algorithm calledfrequent subtree mining algorithm was also given. The experiments were carried on thedata of different domain-specific query interfaces. The results indicate the soundnessof the proposed model and feasibility and effectiveness of the designed algorithm.However, using the above model can not obtain a desired integrated query interfacewhen dealing with the query interfaces with great structural difference.To overcome the shortcomings of the first model, an improve model is also proposed for the integration of query interfaces in this paper. First, the ordered tree of aquery interface was transformed to a structural matrix according to its structure. Ametric of similarity of different ordered trees was proposed. An ideal ordered tree usedas a comparison tree was constructed based on the frequent structure of all queryinterfaces. Then, all the attributes were sorted according to the sequence constraints ofthe query interfaces, and a measurement criterion for this sequencing was given. At last,aiming to optimize the sequences of leaf nodes in the integrated tree and the similaritybetween the integrated ordered tree and the ideal ordered tree, we built a newmulti-objective model for the query interface integration problem. A tree-mergingalgorithm was designed for this model. The experimental results showed the soundnessand effectiveness of the improved model, and also indicated the stability of thedesigned algorithm.