Enhancing Rule-based Ontology Reasoning On Spark

The rule-based OWL reasoning is to compute the deductive closure of an ontology by applying RDF/RDFS and OWL entailment rules,such as TBox and ABox OWL entailment rules,and the reasoning of ontology is one of the key technologies to realize the semantic Web.There is a close relationship between the reasoning efficiency of the rules-based ontology and the order of execution among the rules,and the order of the rules depends on the dependence between the rules.At present,the optimization level of overall rule execution order is not high enough,and the limited computing capacity of the distributed framework resulted in low efficiency of ontology reasoning.Improving the reasoning efficiency of ontology by optimizing the rule execution order has become an urgent problem.In this paper,a local optimal inference scheme is proposed,which makes full use of the dependence among the rules,and we greatly improve the reasoning efficiency of ontology under the Spark framework by further optimize the execution order of the rules.It proposed the complexity of the overall rule execution order based on the dependence among rules,and we decomposed the overall rules into local analysis of rules.According to the data types of conditions and conclusions in inference rules,all rules are divided into four classes,SPO rules,type rules,sameAs rules and schema rules.On this basis,we analyze the dependencies of rules,and establish the corresponding dependency graph.It also gives rule execution order with the highest inference efficiency.We can choose any one rule execution order of each class to form an overall reasoning strategy.Finally,we design the parallel inference algorithm and implement the new rule execution order on Spark in a prototype called RORS.The experimental evaluation is carried out to verify the correctness and efficiency of the proposed scheme.Compared with the current reasoning engine Cichild and KP,our approach improves the reasoning efficiency by 30% on dataset LUBM200(27.6 million triples)and DBpedia(42.5 million triples)on average.In summary,based on the local optimal reasoning model,this paper proposes a new ontology reasoning framework RORS.The experimental results show that RORS significantly improves the reasoning efficiency and has a significant performance advantage.