The Research Of Ontology Generation Method From Relational Database

As a modeling tool which could describe concepts on both of semantic and knowledge layers, ontology could solve two important knowledge engineering problems: knowledge sharing and knowledge reusing. In recent years, ontology has become a research focus in the field of artificial intelligence, which mainly study on the ontology representation, ontology building, ontology integration, ontology application and so on.To reuse and share the knowledge in the existing knowledge system, it's necessary to build domain ontology based on the existing system. At present, most of ontologies are built by the hands of domain experts, which has many disadvantages such as complexity of the project, strongly depending on the experts, too slow for building and so on. To avoid the weakness above, the database resource should be used to build the ontology and in this paper the research of ontology building based on relational database is represented.At the beginning of this paper, a plan of ontology building method is forwarded, which includes four steps: information extraction from relational database, concepts in ontology generation, pointing out the relationship of the concepts, integration with existing ontology. Then the research of concepts generation is described, which is finished by a new idea: firstly generating middle entities from database and then turning middle entities into concepts in the ontology. The middle entity is defined in the paper. Then the generation methods of the middle entities and concepts is studied and discussed. The last two steps of ontology generation belong to ontology integration.Then the methods of ontology integration is studied, which falls into two main steps: using hierarchical cluster method to find similar concepts and using heuristic rule to merging similar discovered concepts. The similarity of concept consists of two parts: linguist similarity and context similarity, which are both studied. Experiments are done to prove this method.At last, a system is explored to realize the method studied in this paper.