摘要: Large graph databases are commonly collected and analyzed in numerous domains. For reasons related to either space efficiency or for privacy protection (e.g., the case of social network graphs), it sometimes makes sense replace original with a summary, which removes certain details about topology. However, this summarization process leaves database owner challenge processing queries that expressed terms graph, but answered using summary. In paper, we propose formal semantics answering on summaries structures. At its core, our formulation is based random worlds model. We show important graph-structure adjacency, degree, eigenvector centrality) can be efficiently closed form these semantics. Further, approach query answering, formulate three novel partitioning/compression problems. develop algorithms finding summary least affects accuracy results, evaluate proposed both real synthetic data.