Efficient Query Answering in Probabilistic RDF Graphs

摘要

In this paper, we tackle the problem of efficiently answering queries on probabilistic RDF data graphs. Specifically, we model RDF data by probabilistic graphs, and an RDF query is equivalent to a search over subgraphs of probabilistic graphs that have high probabilities to match with a given query graph. To efficiently process queries on probabilistic RDF graphs, we propose effective pruning mechanisms, structural and probabilistic pruning. For the structural pruning, we carefully design synopses for vertex/edge labels by considering their distributions and other structural information, in order to improve the pruning power. For the probabilistic pruning, we derive a cost model to guide the pre-computation of probability upper bounds such that the query cost is expected to be low. We construct an index structure that integrates synopses/statistics for structural and probabilistic pruning, and propose an efficient approach to answer queries on probabilistic RDF graph data. The efficiency of our solutions has been verified through extensive experiments.