Large-scale Ontology Structure Analysis Based On Graph Computation

In the field of information science, ontology provides concepts and their mutual relations in shared conceptual areas. Due to the formal representation of ontology can be understood by computer, therefore ontology in areas such as artificial intelligence, pattern recognition, Computer aided translation has become a hot research topic. With the rapid growth of information, the scale of the ontology is also growing. Large scale ontology contains thousands of concepts, relations between the conceptions are complex which makes the operating complexity is high. In order to improve the use efficiency of large scale ontology, how to effectively extract the key information of large scale ontology structure become a hot topic in the study of the ontology.The paper first introduces the commonly method to descript ontology structure, and on this basis this paper discusses the description of ontology structure based on graph theory. Under the background of large scale ontology, ontology concept compares to the vertex of the graph, relationship compares to edges of the graph. Then the paper discusses the graph partitioning strategy based on single multi-core which is the basis of graph computation. Then the paper studies the vertex centric model and the idea of divide and conquer method, we based on PageRank and Label Propagation to design algorithms for analysis of key nodes and segmentation of ontology. In order to verify the feasibility and efficiency of the algorithms. We implement a large-scale ontology structure analysis system based on Graphchi with these algorithms. The experiments on large-scale ontology data set shows that the system can quickly find the key nodes of the structure and reasonable structure division.