Querying RDF dictionaries in compressed space

Abstract

The use of dictionaries is a common practice among those applications performing on huge RDF datasets. It allows long terms occurring in the RDF triples to be replaced by short IDs which reference them. This decision greatly compacts the dataset and mitigates the scalability issues underlying to its management. However, the dictionary size is not negligible and the techniques used for its representation also suffer from scalability limitations. This paper focuses on this scenario by adapting compression techniques for string dictionaries to the case of RDF. We propose a novel technique: D comp , which can be tuned to represent the dictionary in compressed space (22--64%) and to perform basic lookup operations in a few microseconds (1--50μ s ). In addition, we propose D comp as a basis for specific SPARQL query optimizations leveraging its ability for early FILTER resolution.