Research On RDFS Ontology Debugging Using Distributed Computing Technologies

Ontology can be used to achieve some degree of knowledge sharing and reuse. Different ontology language has different expression ability, and contains different level of semantic information. Using the corresponding reasoning engine,we can get the implied information. When ontology reasoning results have logic conflict, we need to debug so as to find and locate the problem. Based on the fact that ontology debugging is complex,it is difficult to debug under large-scale ontology data set. In order to gain the capability of processing large-scale RDFS ontology which is a class of Ontology languages, we use the distributed technology to breakthrough the limitation of I/O and main memory storage.The content of this paper includes:RDFS ontology reasoning under distributed environment, distributed storage and debugging information updates and ontology debugging.During the process of RDFS ontology debugging, much of relevant debugging information need to be accessed frequently, so we need to design the ontology reasoning algorithm to collect debugging information before the ontology debugging. Existing ontology reasoning algorithms involve a large number of iterative arithmetic which seriously reduce the efficiency of reasoning. In order to improve the efficiency of the RDFS ontology reasoning we will use Spark which is a distributed processing framework and more efficient than mapreduce framework when deals with iterative computation. In addition, we use the dictionary coding technology to compress the ontology data, so as to reduce the storage costs and improve efficiency of reasoning.The debugging information is stored in a distributed database system to support the debugging of ontology after reasoning process of. In addition, on ontology updates, debugging information updating algorithm is proposed.Using debugging information gained before, we realize two different algorithms of ontology debugging.The one, first we query the debugging information from the HBase, then find justifications of an RDFS entailment; The other one,we store the ontology data and debugging information in graph data structure of Hama framework, thus from the data structure we also can get the justifications. From what has been discussed above, we implement a prototype system.In this paper, we uses the LUBM data sets for ontology reasoning and ontology debugging experiments. The results show that the ontology reasoning method based on Spark framework achieves better performance than inference engine based on hadoop, and the speed increased by 20～30%. In addition, ontology debugging algorithm can find justifications of an RDFS entailment quickly when dealing with the RDFS ontology which contains billions of triples.