Relational Databases-based Semiautomatic Method Of Ontology Constructing

Ontology can be the basis for knowledge representation and heterogeneous data integration. But manual constructing of ontology is time-consuming and error-prone. And many methods of constructing ontology from relational databases are proposed. The data sources of heterogeneous data integration mainly come from the relational databases. But, the method of constructing ontologies from integrated databases is still rare.In consider of these issues, a semi-automatic method of constructing ontology is proposed in this paper. At first, relation schema file can be extracted from relational databases automatically, and then initial ontology can be automatically constructed from the relation schema file. With the help of initial ontology merging, the object ontology can be constructed semi-automatically. And in this paper, the fallowing tasks will be completed.1. The automatic extracting of relational database schema. At first, an XML database schema is defined. Then we proposed a method which can automatically extract relation schema information from relational databases. And after extracting the relation schema information from relational databases, the relational schema information was written into an XML file.2. The automatic constructing of initial ontology. In this part, the method of how to automatically construct OWL ontology from relational schema files is proposed. The proposed method divided ontology construction process into two parts: semantic finding and ontology mapping. And a graph called M-Graph is proposed. With the help of the generating and analysis of the M-Graph, we can not only mapping the relational schema files into ontologies, but also can find the sematic information that hide in the relational schema files. In the ontology mapping part, combined with some mapping rules and the M-Graph, the relational schema files are automatically mapping into OWL ontologies. 3. The semi-automatic constructing of object ontology. In this paper, the method of constructing object ontologies is to merge initial ontologies and insert some rule into the object ontologies. In consider of the ontology is constructing from relational database and instances can't associate with two concepts, a method of calculating the initial ontology similarity between concepts is proposed. In this method, the sematic similarity, the property set similarity and the structure similarity are all considered while calculating the similarity between concepts. In order to construct completed object ontology, we define the rule that need to be added, and add these rules to object ontology.And a prototype system is implemented. This prototype system can automatically extracting relational schema files from relational databases, and also can automatically constructing initial ontologies from relational schema files. And this system could construct object ontology by mering initial ontology and adding domain rules. An object ontology is constructed from two relational databases which get from hospital domain. This prototype system proved that the proved method can construct domain ontologies semi-automatically. And the result object ontology is extensive and accurate.