To overcome the problems about consequence discontinuities and consequence sorting dissatisfaction of relational database-based key word retrieval, a novel data community-oriented object-level retrieval system is designed and implemented, which is based on data graph. The combination of each keyword nodes that center nodes can reach determines a community. Some corresponding adjustments are carried out by comparing the cost from the central node to each keyword nodes. The implementation of the retrieval system not only makes the retrieval consequences contained more information, but also makes the structure more clearly. The retrieval system achieves a comprehensive, coherent, orderly retrieval for relevant information.