Research And Implementation Of An Ontology Evolution System

Ontology is an important form for knowledge representation, and it is a formal explicit specification of a shared conceptualization. Ontology has been applied widely in many fields such as semantic Web, software engineering and biomedical informatics. With the development of the semantic Web, the distributive and dynamic characteristics of Web make the ontology change constantly, which lead to the ontology evolution and some different versions of ontology. Ontology evolution is the timely adjustment of the ontology and the conforming propagation of the changes. Changes of domain, shared conceptual model and the representation will lead to the ontology evolution. Ontology evolution involves a series of key technologies such as evolution detection, representation of the ontology evolution action, evolutionary strategy, impact of ontology evolution and so on. This thesis discussed the above key issues in ontology evolution.The main contributions of this thesis are listed as follows.(1) Proposing a method for detecting the ontology evolution based on text similarity. Through parsing the ontology elements and some preprocessing such as splitting words, deleting stop-words and so on, we get the Semantic Description Document. And then this thesis utilizes text vector space constructed by TF-IDF method to calculate the similarity of the two elements to determine whether two elements have changed. On this basis we detect the ontology evolution. Choosing OAEI benchmark 2009 data set to validate the method. The experiments demonstrate that this method is effective. But it also remains to be improved in the ontology of which structure is complex, because that the method is based on text similarity, it does not take account of the hierarchical relationship of the ontology and ignores the situation that evolution may occur in hierarchy.(2) Proposing the classification method of ontology change actions and an evolution strategy. Proposing concepts of basic change operations and composite change operations. The basic change operations are complete and can not be decomposed more, and all operations can be made up by basic change operations. On this basis we develop the composite change operations, propose the decision point and an analyzed the operations in the decision point of ontology. Meanwhile the definition of cost is proposed, and this thesis develops the minimum cost evolution strategy.(3) Proposing three ways to measure the ontology evolution impact. On the basis of ontology evolution detection and classification of the evolution actions, this thesis proposes three ways to measure the ontology evolution impact:the basic impact measurement, the semantic impact mcasurement and the measurement based on Gauss function. The basic impact measurement calculate the impact from the perspective of overlapping element. On this basis the semantic impact measurement considers the semantic relevance of the modified element. The measurement based on Gauss function puts forward the correlation formula among ontology elements, and considers the impact of three basic change operations respective ry. The experiments show that all of the three methods can measure the evolutionary trend. And meanwhile the measurement based on Gauss function has more reference meaning, because this measurement considers the relationship among ontology elements and influence degree of different change actions, which make it consistent with the objective laws of ontology evolution.Based on the above work, an ontology evolution system is developed, which is equiped with a series of features such as evolution detection, classification of the actions, evolution impact analysis and so on.