Study On Deep Web Query Interface Pattern Matching And Query Results Annotation

With the rapid development of the Network, The information hidden in theback-end database of Web resources, because of its large amount of data, the integrity ofthe structure has been widespread concern, this information usually requires the user tosubmit queries to access and not for traditional search engines index and thus theformation ofthe Deep Web. Deep Web has become the most important source for peopleto getnetwork information, how to accurately, quickly and easily obtain informationfrom the Deep Web site is becoming the main concern of researchers, so Deep Webinformation integration gradually become a important research questions of the studyarea.Deep Web query interface integration and the Deep Web semantic annotation aretwo of the key questions in the Deep Web information integration. The field ofIntegrated query interface is help to establish a unified query interface which can easilyquery the Deep Web Information, and also is the first step in the Deep Web Integration.Semantic annotation allow the computer to understand the semantics of the resultsinformation, and then displayed to user in a more friendly way.In this paper these twoaspects of the problem were deeply studied andput forward the improvement methodand experimental verification.The main research work as follows.①Paper first introduce the Deep Web information integration system framework,and then pattern matching and semantic annotation of research status is summarized,Finally study and analysis of the shortcomings and deficiencies of the traditional patternmatching methods and semantic annotation methods..②Based on analysis of the concept of similarity calculation algorithm andWordNet Dictionary, a similarity calculation method based on the edge weights betweenconcepts is proposed, which is to solve the problem of the accuracy and applicability ofexisting methods. This method is to give the shortest path between the concept of theirside contains a different amount of information given different weights, and then use thenonlinear equations fusion weighted shortest path and minimum upper father conceptualcontent. Experimental results onM&C dataset show that the new method is able toachieve a higher correlation with manual methods.③Proposed a combination of matching degree and semantic similarity Deep Webquery interface mode matching method, this method in order to solve the existing method of matching efficiency is not high and complex matching problems. The methodused matching degree to measure the relationship between the interface properties,according to the value of the correlation coefficient to determine the combination andthe synonymous two kinds of relationships between query interface attributes. Finallycalculate semantic similarity between synonymous relationship concepts.Theexperimental results show that this method can effectively improve the matchingaccuracy.④In order to accurately and completely mark result records proposed a new queryresult semantic annotation method based on DS evidence theory. The label vocabularyare stored in the DS evidence theory recognition framework, And then under theframework using different methods for labeling, and finally using the improvedsynthesis method synthesis of the results of the various methods.The experiments showthat the new method can be efficient labeling query result records.