Research On Data Integration Based On Ontology Technology

With the continuous development of modern technology,the computer has become an indispensable part of people's lives. In recent years, on the premise of more and more data, the storage of data has become the main problem. Because most of the existed data are made based on different platforms,use different formats and for different purposes,therefore they are featured by heterogeneous and distributivity. This feature greatly affects the effort of interact and data reuse. To solve this problem, scientists all over the world put forward a number of methods for data integration. Caused the method based on the effect of "ontology"concept did particularly well,it is now has become the new hot spot in research.In this thesis, the theoretical basis and related technologies used in ontology-based data integration described at the first.After depict the outline of ontology's theory,system framework of data integration is given.And the various parts of the technique used in the framework are described in detail,including:XML conversion,construct ontology,XML and OWL mapping and query processing.After this,article will focus on the ontology mapping technology which is very critical .As more and more similarity algorithms have been proposed,ontology mapping models also began to diversify, but after careful study, we found that mature models and similarity algorithm that commonly used still have many limitations,Through the full analysis of the characteristics in existed common models,several prevailed shortcomings are found.they are large computing capacity,low degree of automation,difficult portability and single algorithm.In order to improve these four common disadvantages, this paper presents an innovative ontology model called improved multi-strategy hybrid mapping model(O-ESMR ).This model is composed by four key modules which are feature extracting,concept screening, multi-strategy mapping and results dealing.The thesis focus on the innovative modules of concept screening and multi-strategy mapping. In the module of concept screening,by using the technique of WordNet,we can calculate the related degree of words which composing the two concepts.Then conceptual similarity could be reckoned through the related degree.After compared with the threshold,we can screen the candidate sets of concepts.The purpose of this module is to improve the shortcoming of large computing capacity.Multi-strategy Mapping module not only improved the shortcoming of low degree of automation by the way of automatic weighting,but also improved the shortcoming of difficult portability and single algorithm based on multi-strategy hybrid approach including name,properties,structures and examples.Finally, the experiment using multi-strategy hybrid mapping is proposed. Data sets applied for are benchmark data sets which provided by OAEI (Ontology Alignment Evaluation Initiative). The results of this mapping are compared with the mature models.After the full analysis and study of experimental results, it was found that the model reducing the time complexity of the algorithm by the means of reducing the amount of calculation on the premise of precision and recall.