Research On Method Of Ontology-Based Knowledge Integration

In knowledge economy era, with increases of products'function complexity, quality requirements and market reaction speed, the knowledge intensity of enterprises are higher and higher. Since the number of heterogeneous knowledge sources in enterprises has been increased, heterogeneity has been obstacles to knowledge and information sharing and interoperation in enterprises. Thus, integrating knowledge from heterogeneous sources into an integrated knowledge base according to a unified conceptual model, vocabulary, and knowledge representation plays an important role in making the most of heterogeneous knowledge sources, promoting utilization of enterprises'knowledge, and improving the quality and speed of product design. Several key technologies in knowledge integration have been studied in this dissertation. The research and results are as follows.Firstly, the object-oriented concept graph is put forward to represent ontology. An object-oriented concept graph is a directed graph consisting of concept sets and relation sets. In an object-oriented concept graph, concepts are represented in the object-oriented methodology, while relations between any two concepts are represented in conceptual graph. The object-oriented concept graph has advantages of both conceptual graph and object-oriented representation. It can represent the intension of concepts integrally, and can represent relationships among concepts clearly.Secondly, regarding ontology as a unified conceptual model, a knowledge lifecycle oriented knowledge integration process model and architecture are proposed.Thirdly, the approach for ontology construction is proposed. On the basis of an improved semantic affinity based term clustering algorithm, a semantic-based clustering approach for concept generation is proposed. Domain concepts are generated through semantic affinity based term clustering while terms are extracted from term sets of heterogeneous knowledge sources. Then, a knowledge-based method for defining relations between concepts is proposed. The super-concept sub ontology will be constructed in advance and the ontology will be constructed on the basis of the super-concept sub ontology.Fourthly, a Genetic Algorithm based production rule sets integration method is put forward to integrate rule sets. Rules and rule sets are encoded with binary strings. Rule sets'binary strings are regarded as individuals, and fitness function reflecting the correctness and completeness of the rule sets are defined. The best individual is obtained after genetic evolution by SGA. After post treatment and decoding, the integrated rule set can be obtained. An example has been computed by the method. Redundancy, contradiction, and other kinds of inconsistency in initial rule set are eliminated. Compared with other methods, the integrated rule set obtained in this dissertation has higher correctness and completeness.On the basis of above research, an ontology-based knowledge integration prototype system is constructed. Several main modules are developed including ontology construction tool OntoBuilder, Genetic Algorithm based knowledge integration tool GABKIT, and integrated knowledge base management system IKBMS.