Studies On The Method Of Fuzzy Ontology Mapping

The Semantic Web is considered as the extension of the current Web. It can append the machine-readable and machine-processable information to the knowledge in the Web, to achieve the semantic communication between people (machine) and machine. In order to organize the knowledge information in the Semantic Web effectively, ontology is used as the knowledge representation model. However, different ontology-constructors apply different terminology sets and different knowledge structures to build ontologies, and it necessarily leads to the problems of semantic conflict and heterogeneity among ontologies, which seriously influences knowledge sharing and reuse, and semantic interoperability among different ontologies. To resolve the above-mention problems, the method of ontology mapping is applied to create mapping relations between the elements that are the same or similar for the heterogeneous ontologies.With the development of the Semantic Web. fuzzy information is employed to improve the representation of knowledge in many application domains. But, the classic ontologies can not represent this kind of fuzzy information as the limitations of its own. To this end. many researchers introduced fuzzy set theory into classic ontology, and generate a new knowledge model for dealing with fuzzy information:fuzzy ontology. As the classic ontology, semantic mappings between heterogeneous fuzzy ontologies are necessary, in order to resolve the semantic conflict problems generated by heterogeneity. According to the current research progresses in (fuzzy) ontology mapping methods and (fuzzy) concept similarity calculation methods, it is found that:(1) In the current methods of fuzzy ontology mapping, the information content of fuzzy concept is not applied to the similarity calculation for fuzzy concepts. Therefore, these methods do not consider the influence of the membership relation of instance to fuzzy concept on the calculation of information content of fuzzy concept.(2) When computing semantic similarity for fuzzy concepts, the current methods of fuzzy ontology mapping often ignore the implicit information of fuzzy concept. Moreover, these methods do not make use of the existing mapping-relationships reasonably to avoid the problem caused by the double counting of semantic similarity.(3) The current methods of fuzzy ontology mapping take as input two fuzzy ontologies, and create mappings between them. Therefore, they are not fit for creating mappings among multiple fuzzy ontologies in the same domain.For this purpose, studies on the membership-relationship of instance to fuzzy concept, the representation form of fuzzy set for fuzzy concept, the role of implicit information of fuzzy concept, and the system structure of multiple-ontology mapping are presented. The specific research work contains the following aspects:(1) The similarity calculation method of fuzzy concept is provided based on information content. This method firstly analyses the inclusion relationship between the instance set of sup-concept and the instance set of sub-concept, and creates instance sets for all the fuzzy concepts from the bottom up. Then, the calculation formula for the membership of instance to fuzzy concept is given, by taking into account the relationship between the property value of instance and the property domain of fuzzy concept, and the weights of property domains. Next, the information content of fuzzy concept is computed according to its instance set. Finally, the semantic similarity between fuzzy concepts is computed based on the information content.(2) The method of fuzzy ontology mapping is proposed based on association rule, to solve the mapping problem for two fuzzy ontologies. Firstly, the f-RDF triples contained in fuzzy ontology are translated into association rules, and the new association rules are generated by the transitivity of association rule, which can be used to represent the implicit information of fuzzy concept. Then, the association rule sets are generated for fuzzy concepts, and these sets are put into a queue of association rule set ordered by the number of set. Next, the similarity method for association-rule set is given, according to the characteristics of association rule. Finally, an iterative mapping process is applied to create mappings between fuzzy ontologies.(3) The mapping method for multiple fuzzy ontologies is proposed based on conceptual graph, to solve the mapping problem for multiple fuzzy ontologies in the same domain. Firstly, all the fuzzy ontologies in the same domain can be divided into a reference ontology and multiple source ontologies. Then, the reference ontology and source ontologies are separately translated into two sets of conceptual graph:{R-Set} and{S-Sets}. Next, the similarity method for conceptual graph is given by analyzing the structure of conceptual graph. Next, the mappings between reference ontology and source ontologies are created based on the similarity result. Finally, the mappings among source ontologies are created according to the mappings between reference ontology and source ontologies.