Research On Key Technology Of Efficient Semantic Index For Large-Scale RDF Data

With the development of Semantic Web, the data of each research area according to the semantic web data format present a geometric explosive growth. The number of RDF distributed on LOD(Linked Open Data) has reached a scale of ten billions. There are mainly two kinds of work relating to RDF data, namely query and inference. The efficient query work relies on the undercourse index structure; while the inference work are relating to closure-computing according to the inference rules of the Semantic Web, which has a high complexity.Recently, there are plenty of researches on query and inference work on RDF data, of which the main drawbacks are of the following two points. First, the undercourse index of the RDF data storage is structural index without any semantic information; second, the inference work is off-line without supporting the on-line true-time inference. In this paper, by researching the RDFS inference rules and combining them with the ORDPATH encoding schema, we propose a coding schema called Resource Prefix Code which expresses the hierarchical and entailment relations between the resources of RDF data, and propose an semantic index structure for large scale RDF data. ABox and TBox are distinguished in the RDF data, and construct the TBox semantic relations using the ORDPATH encoding schema. Then persist the semantic relations into the RDF SPO index, which loads the SPO index with semantic information. The obvious characters of the novel index are that by just querying the RDF data we will get the knowledge entailed by the RDFS, which realizes the RDFS Entailment Regime by using the hybrid inference method. A series of experiments are designed and conducted to compare the results and performance of the query on semantic index and traditional index.The analysis and experiments in this paper show that, in aspect of the storage space and time overheads of data loading, or in the aspect of the query processing, the semantic index structure supports efficient query along with effective inference without adding obvious overheads to the traditional index, thus realizes the true-time inference of RDF data.