A Research On The Structure Of Ontology Networks And Its Evolution

With excellent characteristics of knowledge sharing and reuse, and also as the core of the semantic web and knowledge organization techniques, ontologies are used more widely in scientific research and practical applications in recent years.However, with the development of Internet and semantic web, the number of ontologies becomes more and more. On the one hand, under the situation of epidemic construction activities of ontologies, each institution or scholar who study ontology or semantic web and other related fileds claim to authority about their own ontologies, but the quality of ontologies is not that good in fact to some extent. On the other hand, the differences between different fields and the development of fields lead to the scale of ontologies become more and more huge, and this brought new challenges to ontology understanding and ontology reuse.At the same time, the change of domain knowledge and user demands causes continuous evolution of ontologies and users need to understand ontology better from a dynamic view. It also makes ontology management activities more diffcult in this situation. Under this environment, this paper uses network analysis techniques to study the complexity and evolution process of ontology from three angles:the complexity of topology structure of ontology, ontology network modularity and ontology network evolution cost. The study deepens the understanding of ontology structure further and it provides techniques and methods for ontology reuse with ontology modularity and it also provides guide for ontology management activities.Through the analysis of the similarities and differences between ontology and traditional study of complex networks, this paper points out that the analysis of ontology networks topology is different with traditional networks analysis. Through designing a new formanization method of ontology, it slao points out that ontology network is a hierarchical network with semantic meaning entailing in itself, and with multiple edges and loops in, and it is a n-type network with k kinds of relations. This paper explores the structure complexity of ontology networks from perspectives such as:ontology vocabularies distribution, network degree distribution, clustering coefficient, length of shortest path and ontology network hierarchies, etc. It finds that OWL ontology vocabulary distribution follows exponential distribution.This paper verifies that large-scale ontology network follows power-law distribution, and it is a scale-free network. But the power-law curve may be associated with turbulence. Ontology network’s clustering coefficient is very small and it doesn’t have the characteristics of small world effect. This paper also finds out that ontology network is a flat spherical hierarchical network, rather than pyramid shape. In ontology network, leaf node concepts account for large proportion, and the inheritance relationships occupie the main position, and most of the nodes only have out-degree without in-degree. The metrics related to network node degree is not the suitable one for the measurement of the field importance of nodes. Nodes with high degree value concentrate in the upper hierarchy of ontology. Clustering coefficient-degree distribution of ontology network follows scale-free characteristic, and this veiries the hierarchical characteristic of ontology network. The length of shortest path distribution increases firstly and then decreases lastly. This paper has designed a method for calculating the ratio depth of ontology concept node, and it finds out that most non-inheritance relations lie in the bottom of ratio depth and the middle layer of ontology hierarchy. Along with the increasing of ontology level, the node number with the same degree increases at first and then decreases at last, and level distribution curve of different node degree is similar to each other.Research perspective from ontology modules is larger than from ontology concept nodes. In the ontology modularization study, this paper analysises the inadequacy of traditional ontology modularization methods and network community detection methods used in the ontology modularization. Module partition results are not comparable with these methods due to different applications, and they rarely consider the semantic characteristic of ontoloyg networks in actual ontology modularization process. This paper designs an ontology modularization method combing with ontology topology structure and its semantic meaning together, and this method is applicable to the reuse of ontologies, ontology reasoning and ontology visualization in multiple scenarios. At the same time, the method has a strong flexibility and it can produce ontology modules of different scale according to the actual application scenario.Based on the Gene Ontology, this paper carries out modularization operation using this new method, and verifies the effect and significance of ontology modularization from the module number, module scale, module cohesion and coupling degree perspectives in ontology modules visualization application. It also finds that the ongology modules have the charactics of self similarity, and smaller modules’shapes are closer to star structure. Ontology modularization makes segmentation, reuse and visualization of large-scale ontology possible. It is more macro from the perspective of ontology evolution to observe ontology. In the research of ontology network evolution, this paper explores ontology evolution’s motive and influence, and the role of ontology evolution research. Ontology evolution is a process with a complex life cycle, and it has great difference with traditional network evolution model. There is not a unified evolutionary random model of ontology network. This paper summarizes the ontology change operation strategies and sub-change operation strategies, and formanize them, then map them to nodes and arcs operation of ontology network. Through analyzing the change operation strategies in GO, it further finds that the ontology evolution operations are complex.Different ontology evolution operations will cause different diffusion effects, and different needs will produce very different sub evolution operation strategies. There is no uniform sub evolution operation execution path in ontology evolution process. Influence of ontology evolution on specific application is enormous. For example, ontology evolution can cause revision of semantic annotation results which are based on ontology. And it also causes ontology knowledge base need to be modified, for instance, if a concept in ontology is removed, then all instances of the concept need to be deleted or re-distributed. This will introduce a huge impact on ontology knowledge base retrieval and application.This paper constructs a cost model of ontology evolution which consists of ontology structure cost and application cost to measure ontology evolution operation cost. A minimum cost algorithm of ontology evolution is designed which can calculate evolution path under the minimum cost according to evolution demands. Scientific validation of the model is carried out by the Gene Ontology that illustrates the validity and superiority of this model. Ontology evolution cost research has significance on ontology management.