Research And Implementation Of Ontology Matching Approach SGOM

Ontology is a Conceptual Model which is used for describing information systems on the level of semantics and knowledge and it has been used in many fields. There are differences between Ontologies because of different builders of Ontology modeling and different ways of Ontology modeling. The Ontology matching has become the core of Ontology study.The purpose of Ontology matching is to reuse the existed Ontologies and combine and unfold them through certain methods. Until now, the achievements of automatic or semi-automatic Ontology matching are very limited. There are many challenges in this research field, and there is a lot of work to be done in automatic Ontology and semi-automatic Ontology matching.Currently, the development of the Ontology matching is still imperfect. Ontology matching needs further improvements, even a totally new method should be proposed to realize fully automated of Ontology matching. Therefore, it is meaningful to propose a method without manual interference and with a good recall and precision. Only when the problem of Ontology matching is solved, Ontology can be widely used.Actually, in order to improve the accuracy of Ontology matching, several methods and technologies are combined and put into use other than a single way. For these reasons, this paper proposes an Ontology matching method-SGOM. It combines structure-based method and name-based method to obtain better precision and better recall. SGOM has element-level matching, as well as structure-level matching.There are two name matching methods to compute the similarity in the name matching method of SGOM. SGOM computes the name similarity by computing semantic distance of name. The name similarity computed by this way can reflect the semantic relationship between names more accurately. The computation of semantic similarity uses the language Ontology library WordNet. SGOM name matching method calculates the semantic similarity of every matching pairs (x, y)∈(CST1×CST2), where x and y are root sequence of the name. Because the name can be made of several words'abbreviations, the words in the root sequence may be not the English words. Under this circumstance, the semantic similarity of matching pair(x, y) is zero. Then the SGOM confirms the name similarity through calculating the editing distance of the matching pair(x, y) to make up the calculating deficiency of semantic similarity in name matching method.The name matching of SGOM makes up for the disadvantage of traditional methods of name matching, that is, the similarity of the Morphology can't reflect the similarity of the sense of a word. The names being matched by matching method of SGOM method are Ontology elements which must come from the same sets, for instance, matching between the class name and the class name, matching between the property name and the property name. This way of matching minimizes the unnecessary computation, and then reduces the time of computation.The structure matching method of SGOM extracts the ontology's structure from the OWL Lite file. The Ontology presentation language OWL Lite has abundant semantic information. SGOM implements the structure matching of Ontology by using the abundant Ontologies semantic information from OWL Lite language. SGOM implements the Ontology matching structure on semantic level by using ontology structure and semantics provided by OWL Lite.By combining these two methods mentioned above, SGOM increases the similarity of the ontology elements which have low similarity of name itself but have high similarity of the essence; meanwhile, SGOM can decrease the similarity of the ontology elements which have low similarity of the essence but have high similarity of name itself. So the matching result of the ontology can be improved.The first step of the SGOM method: Ontology Information Extraction. Extract the Ontology information described by OWL Lite files; build the graphic structure of Ontology information. The graphic structure of Ontology information fully makes use of the Ontology semantic information in the OWL Lite files and provides the Ontologies structures with abundant semantic information.The second step of the SGOM method: Name matching. SGOM extracts names of Ontologies elements from the graphic structures of Ontologies information; do match by using of the names of Ontologies elements; produce two similarity matrixes. The name matching method does match only between elements from the equivalent sets. The name-based method combines the linguistic-based method and string-based method, and it greatly improves the accuracy of similarity in the elements-level matching.The third step of SGOM method: structure-based matching. According to different semantic relationships, ontology graphic structure is changed into three Change Weights Semantic Graphs. SGOM sets the initial weights for the different directed edge in the Change Weights Semantic Graphs, takes the similarity matrixes produced by the second step as the initial values, and computes the similarities in the different Change Weights Semantic Graphs iteratively. After each computation, the weights of directed edge in Change Weights Semantic Graphs will be reduced until zero. The method finishes when the weights of all the directed edges are zero.The forth step of SGOM method: SGOM sets the strategy of choosing, and chooses some results from the similarity matrix as the final matching result.The focus is"changing the weights"in the structure-based method of SGOM method. In the process of iterative computation, if the weights of the directed edges in the Changing Weights Semantic Graph are changed, that means the structure of Ontology is changed. This guarantees that the structure of Ontology makes an appropriate effort in the whole method. The commutating times are limited in the SGOM method because that the weights will be zero definitely. So SGOM doesn't have to make sure that there are no circles in the structure of Ontology. This broadens the scope of the Ontologies which are prepared to be matched.Experiments show that SGOM only match between the elements from the equivalent sets. SGOM reduces the numbers of the matching pairs, reduces the redundancy matching and reduces the calculating time. SGOM method makes use of the semantic of Ontology presentation language OWL Lite and Ontology's structures, and improves the efficiency of the Ontology matching a lot. SGOM effectively implements the Ontology matching by using the semantic information of the Ontology, and achieves a satisfying result.Based on SGOM method, the paper introduced other assistant modules and implemented the Ontology matching system, SGOMS. SGOMS takes the OWL Lite language as its supporting Ontology presentation language. SGOMS takes the OWL lite language as the inputs, and then dose the matching between two Ontology O₁ and O₂ as its result.In this paper, Ontology matching system SGOMS is implemented in the Windows environment, on the Eclipse platform, and with the help of Jena and the kit of JWNL.