Research Of Large-scale RDF Graph Parallel Reasoning Method Based On The MapReduce

With the further study on semantic web, the slowly speed of reasoning on largescale RDF graph semantic is becoming more and more evidently. When the traditionalreasoning method met bottleneck, it is a new approach to realize parallel reasoning onlarge scale RDF graph with mapreduce and HBase. Parallel reasoning of RDF graphinvolves how to break RDF graphs to RDF triples, data storage, data reasoning andquery. In this paper, we will optimize the RDF data parallel reasoning mechanismwith mapreduce and HBase. This paper mainly carries the research and design on thefollowing four aspects:First, decomposition of RDF graphic. The original RDF decomposition methodcould not eliminate semantic ambiguity and RDF triples lack of arrangementrelationship. In this paper, we will extend the original RDF decomposition method toeliminate the defects.Second, design of RDF data storage model. RDF data storage is one of the keyfactors related to RDFS reasoning. In this paper we designed three kinds of storagetable which are S-PO, O-SP and SO-P to solve the problem of large-scale RDF datastorage.Third, research and design on RDF data parallel reasoning method. On thefoundation rules of RDFS, analyze the existing RDFS reasoning algorithms, then putforward the defects of original RDFS reasoning algorithm and solve them.Last, apply the theory of this paper in analyzing the social security data anddesign a platform for analyzing the social security data. Then we will test theefficiency and success rate about the RDF graph decomposition and RDFS reasoningon this platform. Finally, we put forward the deficiency of the paper and researchdirection in the future.