Research On Context-driven Knowledge Reconstruction Based On Ontology

To solve a specific domain problem is often difficult to completely rely on a single knowledge source, but on the coordination of multiple knowledge sources. This is because sometimes a single knowledge source can not solve the problem, and multiple knowledge sources can work together to provide appropriate solution. Therefore, domain experts turn to integrating different knowledge sources in order to build a knowledge object which can solve the domain problem.The objective of knowledge integration is to generate a comprehensive knowledge base.Because the integration of whole knowledge sources will increase the complexity of the interoperability among those knowledge sources, therefore, in order to obtain the comprehensive knowledge object which can meet the domain requirements, meanwhile to control the scale of the object, the heterogeneous knowledge sources should be intergrated selectively which is based on specific context. All these knowledge components related to specific context should be integrated to a comprehensive view.The goal of context-driven knowledge reconstruction is to build a comprehensive knowledge base by integrating some related knowledge objects according to a specific context.Knowledge reconstruction is not simply to integrate knowledge sources in those entirety, but firstly to identify related knowledge components according to a specific context, and then to integrate these knowledge components into a comprehensive knowledge base which can provide total knowledge to solve given domain problem.This thesis firstly introduced the necessity of research on the knowledge reconstruction and related researches,and then studied deeply the model of context-driven knowledge reconstruction, the flexible method of extracting sub-ontology based on the traversal, the dynamic method of ontology mapping. Finally the thesis designed the prototype of this model and conducted the experimental analysis. The experimental results validated that the model and the related algorithms proposed have effectiveness and application value.The main works and contributions of this thesis are as follows:(1) Proposed the model of CKReC(Context-driven Knowledge Reconstruction),which is based on the Semantic Web Framework and requires the input of knowledge sources modeled by ontologies. The main steps include defining the problem context, extracting the sub-ontology, ontology mapping, detecting and eliminating the inconsistencies. CKReC not only strictly controlled the scale of reconstructed knowledge, also provided the reasoning ability based on reconstructed knowledge.(2) Proposed a flexible method of Sub-Ontology Extraction based on Traversal(SOET). This method supports users to select multiple classes, attributes and individuals as context of extraction algorithm. The sub-ontology obtained can not only satisfy the domain requirements, also avoid the unnecessary to expand the scale of sub-ontology. The method allows the user to predefine the boundary threshold to control the scale of the sub-ontology flexibly.(3) Proposed a method of context-driven dynamic ontology mapping(CDOM).This method is based on user's predefined problem context, and works effectively in the absence of predefined mapping.(4) Proposed a method of complex entity oriented mapping(CEM). In this method, not only the atomic mapping which only involve simple entity but also the complex mapping which involve complex entity can be found.This mapping method is still effective in the absence of individual.