Abstract: RDF(Resource Description Framework) has been widely used for semantic representation and processing of big data.Traditional RDF can only represent static semantics and can not meet the needs of processing semantics dynamically over time in time-sensitive scenarios.Therefore,many temporal RDF models are proposed,including RDF model for transaction time,RDF model for valid time,and bi-temporal RDF model thatsupports both transaction time and valid time.To support efficient proces-sing of large-scale temporal RDF data,this paper proposes a three-level index structure based on bi-temporal RDF model.Specifi-cally,in the first level of this index structure,the dataset is divided into different subsets according to the update times of the temporal RDF data.In the second level,a quadtree is built for indexing time information in each subset,and in the third level,the bitmap with three composite keys is used to index the subject,predicate,object of RDF triples.Experiments are conducted from three aspects:the time of building index,the index size,and the required query time.Experimental results show that the proposed indexing scheme can reduce the query time effectively and improve the query performance.

Key words: Bitmap index, Quadtree, RDF, Temporal data, Three-level index