Research Of Heterogeneous Data Integration Based On Ontology

In recent years, with the rapid development of Internet and the growing of enterprise informationization, a variety of application systems are extensively applied in enterprises. The datasources of these systems are heterogeneous since these systems were developed in different time and by different department, and they are separated and enclosed with each other frequently. And then, it's hard to share and merge together among these datasources. With the strength demand for obtaining complete, distributed and heterogeneous information, data integration emerged. Data integration aims at achieving exchanging and sharing of those datasources, shielding the heterogeneity between them, and providing a uniform view for users to manage data from different datasource.This paper mainly studies the heterogeneous data integration based on ontology, and brings foward system architecture on the basis of summary of current data integration methods and its key technologies. The main researches and achievements are as follows:①Analyse the existed problems in current data integration methods, and introduce some of classic data integration architectures. Put foward the architecture of information integration based on ontology, and give a detailed description for each part. This framework is loose coupling, expandable and giving support to semantic query.②Study the key technologies: ontology building, ontology mapping and query processing.③Ontology building. Study the method of building global ontology, and local ontology. Local ontologies are built through cramping out semanteme from data resource including structured data(relational database), semi-structured data(XML files) and unstructured data(WEB data, text data). Global ontology needs domain experts' help.④Ontology mapping. Study the mappings between global ontologies and local ontologies and mappings between local ontologies and data resources. Get mappings between global ontologies and local ontologies by using the multi-strategies method based on learning and HowNet. Get mappings between local ontologies and data resources in the building process of local ontologies.⑤Query processing. Put foward a new language called LSQL as global query language. Query processing contains parsing of global query, binding of query variable, decomposition of query and rewriting of query. Parsing of global query generates a query tree. Results of query are returned with a uniform form to users. Advance the class-source mapping table to reduce query decomposition and achieve query optimization.⑥Give the significance of the research and the future work at last.