| With the development of digital technology, a large amount of digital information with heterogeneity and heterogeneous fields obtained from various approaches and geographical distribution is rapidly growing exponentially. How to effectively find out the explicit and implicit knowledge existing in these data resources and realize the collaboration and resource share for the services in modern knowledge is an important research topic. Currently, the research of knowledge grid has been rapidly developed. The closely related ontology is an important tool for knowledge reusing, sharing and modeling, which becomes one of the important research fields. So far, we have established a series of engineering methods for building ontology with a number of theories, techniques, representation languages and tools. However, these theories and methods are mostly based on the hand-built ontology, which is a time-consuming, laborious, onerous, and difficult task, and is prone to bias mistakes. It is difficult to update timely and dynamically. We have studied how to access to domain knowledge and build or expand ontology in the semi-automatic/automatic ways from the existing sources, including text, dictionaries, legacy knowledge database, and internet documentation, which is an ideal and effective way to develop the ontology. The utilization of KDD technology designates the specific direction of the large-scale ontology and application.In the dissertation, first, the theories of domain ontology and construction were introduced, and the theme and method of building domain ontology was proposed. According to this way, from the rich framework ontology to the complete ontology, every step in the rich and expansion with the guidance of ontology are gradually completed to reduce human participation. Meanwhile, a new strategy was proposed to construct the domain ontology framework based on construction of domain concept system. The dissertation illustrates in detail the automatic construction of the domain ontology framework through the information extraction, the modeling and the concept of hierarchical theory.Second, to expand the field of ontology concepts by utilizing the clustering and classification technology in KDD was also proposed. Due to the limitations of traditional concept clustering for the certain domain concept clustering, a domain concept clustering method is proposed, which builds precise domain concept clustering based on rough conception clustering stage. This method first gains the domain concept by extending index words in text clustering and processing roughly concept clustering based on the improved affinity propagation (AP-SVM) clustering, and then the hierarchy relationship in the precise concept clustering is defined under the guide of domain ontology.Third, the present research comes up with an approach to extend the ontology rules by utilizing the technology of multidimensional association rule. The conception ontology is enriched and extended by the ontology rule extraction, consistency treatment, rule mapping establishment, and the re-identification and the renewal of conception ontology. The experimental results show that the proposed approach can be readily implemented and possess higher levels of feasibility and validity. Finally, a knowledge reuse method based on knowledge equivalent mapping is proposed to solve an issue on existing heterogeneous knowledge reuse from knowledge representation level, since the existing knowledge representation and ontology representation are not dependent on the same logic system. This approach realizes auto-reuse of existing knowledge in the course of ontology construction by using the semantic equivalent extraction, the consistency treatment, and semantic mapping. The experimental results show that the approach possesses relatively high accuracy, feasibility, and validity. The research content for the domain ontology construction has great significance in both theoretical guidance and practical application.
|
PDF Full Text Request