Ontology Building Heterogeneous Database And Mapping Study

With the emergence and development of the Internet, it is possible for the exchanging and sharing of information between industries and departments. However, with the development of society, information sharing need to be improve to a higher level to meet people's need, to eliminate the confliction and abnormity of data in heterogeneous database, with which the information integration can be achieved. The key and difficulty of information integration for heterogeneous database are how to resolve the semantic heterogeneous problems. Ontology can express the concept of certain areas and the relationship between them clearly, so integrating information based on ontology can solve this problem very well.The first step of Ontology-based semantic integration of heterogeneous database is to build ontology. The quality of ontology has a direct impact on further application and research. At present, most ontologies are built manually by the experts in the field, which have many disadvantages such as complex process, slow building and so on. Building ontologies for heterogeneous database is to use the existing information resources to construct ontology. The paper makes a deep research on building ontology based on relational database. By analyzing the relational schema of the relational database, we establish a series of rules for getting basic elements for building ontology from relationship schema and put forward a framework for building ontology from relational database. The First step of the framework is to get relational schema from the relational database and recover the lost foreign key, the next step is to get the basic elements of building ontology through that rules, and the final step is to build the local ontology through Jena API.Ontology mapping is the most important step of ontology integration. There are many semantic conflicts between different ontologies, and how to solve them better is the goal of ontology mapping. Chapter 4 of this paper make a specific research on this problem and build a framework to map ontologies with multi-tactic. Ontology elements such as Names, structures and examples have a certain relationship one another, but almost methods have not pay attention to the relationship between the attributes of concepts. So in the framework, we put forward the algorithm of finding mapping of concepts and attributes. Through calculating the similarity of names, structure, examples, attributes of the concepts, we can find out the mapping of concepts between different ontologies. Through calculating the similarity of names, structure, examples, restrains, we can find out the mapping of attributes between different ontologies. In the framework, we build a mapping database to store mapping and similarity of the name which appeared in the mapping process.In the mapping process, we first standardize the candidate of mapping, then we search the mapping related table. If we can't find the similarity of the words, we use the approach which calculates the semantic similarity through WordNet with searching similarity from statistics database on web or measure semantic distance between words. On the basis of the words similarity, we calculate the structure, instances, attributes of the concepts and structure, instances and restraints of the attributes, in which those with high similarity can be recognized as success. Finally we optimize the ontology through experts' participation, which can find the incorrect mapping and search the potential mappings, and implement the mapping process for the local ontology. The mapping database play an active role in calculating similarities, and it can also reduce the time in calculating saimilarity of words, which increase the speed of mapping greatly.We use the rules to build two experimental ontologies through analyzing two relational databases of school. We also make details on the mapping process and point out the key problems of it. From the results of mapping in many experiments, we find that the precision and recall is higher when we add attributes' similarity to the results, especially when the elements of the ontologies are abbreviated. At the same time, by extending the radius of semantic can help to improve the precision and recall too.