The growing volume of RDF documents and their inter-linking raise a challenge on the storage and transferring of such documents. One solution to this problem is to reduce the size of RDF documents via compression. Existing approaches either applywell-known generic compression technologies but seldom exploit the graph structure of RDF documents.Or, they focus on minimized compact serial isations leaving the graph nature inexplicit, which leads obstacles for further applying higher level compression techniques. In this paper we propose graph pattern based technologies, which on the one hand can reduce the numbers of triples in RDF documents and on the other hand can serial ise RDF graph in a data pattern based way, which can deal with syntactic redundancies which are not eliminable to existing techniques. Evaluation on real world datasets shows that our approach can substantially reduce the size of RDF documents by complementing the abilities of existing approaches. Furthermore, the evaluation results on rule mining operations show the potentials of the proposed serialisation format in supporting efficient data access.