Hierarchical System Building Of Wikipedia Knowledge

Knowledge bases in a structured form describe the entities and the relationship between entities in the real world, knowledge representation is the basis of knowledge acquisition and application, through the whole process of building and application of knowledge bases. However, in many ways of knowledge representation, hierarchical knowledge representation has been widely used for its features of simple and intuitive structure, clear semantic relations between entities. Wikipedia knowledge base has the characteristics of high accuracy, wide coverage, real-time updates, Wikipedia link knowledge which constituted by terms and internal links between them owns nodes with high accuracy and edges containing rich semantic information. Based on the consideration of applying the hierarchical knowledge representation method to Wikipedia knowledge, the dissertation proposes some methods to build hierarchical system of Wikipedia link knowledge, and finally in the clear and intuitive way to display the semantic relationship between knowledge.Firstly, the dissertation introduces the main knowledge base-Wikipedia knowledge base, including the basic information of Wikipedia, the knowledge acquisition and construction method of Wikipedia link knowledge, then proposes a solution how to get hierarchical structure from large-scale knowledge link graph. Initially, Wikipedia link graph is seen as complex network community for its uneven features, so related community detecting algorithm is applied to get a series of link sub-graphs based on specific topics, and then use several key nodes for constructing semantic vector space model for each sub-graph. In order to build the hierarchical knowledge system for each link sub-graph based on a particular topic, the dissertation choose nodes with greater Page Rank weight value extending out along edges to discover the closely connected nodes to form clusters. That clusters should be used as semantic units in hierarchical structure of knowledge, so as to build hierarchical structures. Finally, the dissertation adopts agglomerative hierarchical clustering idea, by selecting the appropriate inter-cluster similarity calculation method and the aggregation policy to build the hierarchical knowledge representation.In order to verify the correctness of proposed theory and validity of algorithms, a series of experiments covering all stages of hierarchical system building of Wikipedia link knowledge were conducted. The results verify the feasibility of the theory and algorithms and effectively implement of the hierarchical system building of Wikipedia knowledge.