Research On Integration Technology Of Heterogeneous Data Based On Ontology

Nowadays, information technology is in high developmental speed, a lot of information systemhas been widely used. However, by the reason of creating information platform with astage, purposiveness and dispersive, the data heterogeneity is produced. Due to the heterogeneous dataproblems, communication between various information systems is difficult, and it is very difficult thatinformation is effectively shared, leading to the phenomenon of "isolated is land of information"becoming more and more seriously. So, in order to adjust for the rapidly developing trend of theinformation society, and make the information can be used effectively, we need to design anintegration for the heterogeneous data, at this stage the main task to realize the integration ofheterogeneous data is to solve the problem of semantic heterogeneity.This paper presents a multi-strategy method based on hybrid ontology similarity aiming at theproblem of semantic heterogeneity. Firstly, the paper describes the concept, the type and the target ofthe heterogeneous data integration, summarizes and analyzes the good and bad aspects of severalexisting integration methods, introduces the key technology and function of ontology and ontologymapping in detail, and designs the general framework of the integrated s ystem, introduces thisintegrated system from the user application layer, the intermediate integration layer and theheterogeneous data layer.Then, the paper studies on ontology mapping techniques in the data integration, and focusing onthe research of the method about similarity calculation in the ontology mapping technology. Throughthe analysis of the existing algorithm about ontology mapping technology, the author found thesealgorithms exist some problems, just like large calculation, single algorithms, low level automation andpoor universal. Aiming at these problems, this paper presents the W-NPSI mapping system, this systemincludes the concept feature extraction module, concept set filter module, multi-strategy mappingmodule and result processing module: the concept set screening module presents the calculation ofcorrelation algorithm based on WordNet concept, and set filter module calculates the relationship ofwords based on the position of words in the WordNet, and get the concept similarity, finally screens candidate set of concepts, resolving the problem of big calculation; the multi-strategy mapping moduledesigns an adaptive similarity polymerization reactor, and the central idea is used the adaptive methodto improve the degree of automation system; the multi-strategy mapping module presents a hybridfeature similarity multi-strategy algorithm which includes the concept name, the attribute, the structureand the instance, this method can effectively improves the mapping effect, improves the general of thissystem, and solves the problem of single algorithm.At the end, using benchmark data provided by OAEI (Ontology Alignment Evaluation Initiative)to evaluate the multi-strategy based on mapping algorithm. We could analyse from the experiment thatthis algorithm could guarantee the Full rate and the Precision rate, and at the mean time, reducing theamount of calculation, reducing the time complexity and the space complexity.