Research On The Benchmark For OWL-based Knowledge Base System

Ontology-based knowledge base system is one of important solutions on realizing the high performance storage, query and reasoning of the large-scale semantic web data. At present, there exits a large performance differences among currently available ontology-based knowledge base systems. Thus, benchmarks are developed to evaluate performance for these systems. There are a few problems existing in current benchmarks, such as uneven supports for related technical specifications, poor universality on generation method of instance data, the limited usable range of SPARQL query set, and the lack of support for SWRL inference performance test and so on. Through the analysis of related research on the existing knowledge base system benchmarks, this thesis has based on the recommendation of the world wide web consortium standards such as OWL 2, SPARQL and SWRL, then do research on the OWL-based knowledge base system benchmark, attempting to partially make up the shortage existing in the related research and application fields.In this thesis, the main research work includes several aspects, such as the generation method of OWL ontology instance data, SPARQL query prototype definition and the generation method of SWRL inference rule.Firstly, against the research of the OWL ontology instance data generation method, this thesis has based on the formal syntax analysis of OWL axiom expression, parsing the domain and range of the object properties and data properties in the OWL ontology, then generating batch instances and establishing effective links between instances, so as to realize the generation of OWL ontology data.Secondly, to the definition of the SPARQL query prototype, this thesis firstly analyze the design factors in the existing benchmarks, then put forward the design rules of the SPARQL query prototype. At the same time, this thesis accomplishes the design of the SPARQL query prototypes and the discourse of the SPARQL query generation method.Thirdly, for the research on the SWRL rule generation method, based on the analysis of the mapping relationships between OWL 2 DL constructors and description logic and the analysis of the mapping relationships between SWRL atom and description logic, this thesis discusses the description mechanism of the complex concept in OWL 2 DL, through defining equivalent class axioms for OWL named class, and then parsing the equivalent class axioms to extract semantic equivalence SWRL atom collection, so as to generate SWRL rules.Based on above works, this thesis has designed a set of test programs, which are used to evaluate the performance for the OWL-based knowledge base system. What's more, this thesis has discussed related tests for the accomplished benchmark prototype, and has analyzed the test results. The experiment has showed that the benchmark solution proposed by this thesis is practicable and effective.