%0 Conference Paper
%B 5th European Semantic Web Conference (ESWC2008)
%D 2008
%T Graph Summaries for Subgraph Frequency Estimation
%A Angela Maduko
%A Kemafor Anyanwu
%A Paul Schliekelman
%A Amit Sheth
%K Data summaries
%K Frequency estimation
%K Graph summaries
%K RDF graph patterns
%K RDF query optimizer
%X A fundamental problem related to graph structured databases is searching for substructures. One issue with respect to optimizing such searches is the ability to estimate the frequency of substructures within a query graph. In this work, we present and evaluate two techniques for estimating the frequency of subgraphs from a summary of the data graph. In the first technique, we assume that edge occurrences on edge sequences are position independent and summarize only the most informative dependencies. In the second technique, we prune small subgraphs using a valuation scheme that blends information about their importance and estimation power. In both techniques, we assume conditional independence to estimate the frequencies of larger subgraphs. We validate the effectiveness of our techniques through experiments on real and synthetic datasets.
%B 5th European Semantic Web Conference (ESWC2008)
%C Tenerife, Spain
%8 06/2008
%G eng
%0 Conference Paper
%D 2007
%T Estimating the Cardinality of RDF Graph Patterns
%A Angela Maduko
%A Paul Schliekelman
%A Kemafor Anyanwu
%A Amit Sheth
%K RDF and RDF graph patterns and RDF Semantic Summary and RDF Structural Summary and RDF Query processing and Statistical Summaries and Pattern Cardinality Estimation
%X Most RDF query languages allow for graph structure search through a conjunction of triples which is typically processed using join operations. A key factor in optimizing joins is determining the join order which depends on the expected cardinality of intermediate results. This work proposes a pattern-based summarization framework for estimating the cardinality of RDF graph patterns. We present experiments on real world and synthetic datasets which confirm the feasibility of our approach.
%I 16th World Wide Web Conference (WWW2007)
%G eng