Research On Owl Ontology Modeling Using Relational Databases

The Semantic Web is a new standard of Internet. To realize the semantic Web, a number of domain ontologies are needed. It is an important way of constructing the semantic Web ontologies for ones to translate a large number of the data in databases on the Web into the format of ontology. Such a manual translation is time-consuming and many errors may arise. By means of the rules of mapping between databases and ontologies, databases can be transformed into the ontologies automatically. It is very important for the quick increase of the ontologies in the semantic Web. The techniques of ontology learning based on structural data can get ontologies from databases automatically. In the ontology learning based on relational databases, ontology modeling of relational schema is the first step. But the existing approaches have the following problems:one is that the semantics in relational schema cannot be expressed adequately; another one is that the mapping from relational schema to the ontologies is considered only, and the semantics-preserving migration of tuples receives few attention.This thesis proposes two complementary approaches that can translate relational databases to the ontologies, namely, semantics-analyzing translation and directly-describing translation. For the former, the relational schema is fetched firstly from the relational database and is further recovered to EER model utilizing the technology of reverse engineering. Then the semantic elements in the EER model are mapped into the elements in the OWL. For the semantic elements in the relational schema that cannot be represented directly by the OWL, some semantic elements are particularly defined in the OWL. On the basis, the tuples are translated, in which the semantics of both participation and attributes of tuples are considered. For the directly-describing translation, the structure of relational databases is described and the multiple levels of ontologies are created. The first level of ontologies is about the abstract of relational database. The second and third levels of ontologies are about the relational schema and tuples, respectively. In these processes, the data integrity of relational database are preserved as more as possible by describing the constraints existed in the relational database structure. The two approaches proposed in the thesis consider both the translation of relational schema and tuples. Finally, the thesis develops the prototype that validates the rules given in these approaches.