Research On The Key Reasoning Technology For Large-scale DL Ontology

With the appearance of Semantic Web and the further development of semantic technology,there has been an increasing number of large ontologies in application.Since ontologies suffer from an inherent lack of structure,they are often treated as monolithic objects by the tableau-based description logic reasoning systems.With the increase in scale of ontologies,there has been a declining trend in the efficiency of reasoning systems.Although there exist some efficient reasoning algorithms tailed to specific sublanguages,these algorithms can not guarantee the complete for expressive ontologies.In addition,the standard description logic reasoning systems treat ontologies as static objects.They often repeat the whole reasoning procedure even if there is a small change in an ontology,which causes unnecessary recomputations and leads inefficiency.There is consequently a pressing need for studying the modularity and dynamics of ontologies and optimizing reasoning algorithms.In order to address the issues in reasoning procedure for large ontologies,this thesis presents three research works as follows:(1)Modular decomposition of an ontology.To conquer the limitations of existing decomposition methods,this thesis proposes a hybrid decomposition method for expressive ontologies.First,the EL part of an ontology is encoded into a direct graph to obtain the partial decomposition through reachability and strong connectivity.Then the remaining non-EL axioms are incorporated into previous computation through traditional module extraction.(2)Modular reasoning.Firstly,an ontology is decomposed into several independent modules in term with its modular structure.Then every module is delegated to an appropriate reasoning system.A pay-asyou-go reasoning procedure is implemented in the combined reasoner ComR and a good performance is obtained.(3)Incremental reasoning.This thesis presents an incremental reasoning technology for OWL 2 QL ontologies.Firstly,an OWL 2 QL ontology is represented by a direct graph,then the problem of ontology classification is reduced to the problem of transitive closure.Then an evolving OWL 2 QL ontology is mapping to a dynamic direct graph and the incremental reasoning is obtained by dynamically maintaining the transitive closure.The experiment results shows that the proposed methods are able to cope some issues mentioned above.First,the proposed hybrid decomposition method is able to decompose expressive ontologies efficiently and an average speedup of 6.7-fold is achieved compared to the traditional approach for all the test ontologies.This research result provides modular reasoning with theory and technology support.Second,in the combined reasoner ComR that combines an OWL 2 reasoner with EL reasoner,a reasonable strategy for task decomposition is implemented,consequently a significant speedup is obtained compared with existing modular reasoning method.Third,the proposed digraph-based technology is able to handle evolving ontologies efficiently.Compared with the modulebased incremental reasoning technology,our method is more suitable for evolving OWL2 QL ontologies.