The number and volume of semantic data have grown impressively over the last decade, promoting compression as an essential tool for RDF preservation, sharing and management. In contrast to universal compressors, RDF compression techniques are able to detect and exploit specific forms of redundancy in RDF data. Thus, state-of-the-art RDF compressors excel at exploiting syntactic and semantic redundancies, i.e., repetitions in the serialization format and information that can be inferred implicitly. However, little attention has been paid to the existence of structural patterns within the RDF dataset; i.e. structural redundancy.

In this paper, we analyze structural regularities in real-world datasets, and show three schema-based sources of redundancies that underpin the schema-relaxed nature of RDF. Then, we propose RDF-Tr (RDF Triples Reorganizer), a preprocessing technique that discovers and removes this kind of redundancy before the RDF dataset is effectively compressed. In particular, RDF-Tr groups subjects that are described by the same predicates, and locally re-codes the objects related to these predicates. Finally, we integrate RDF-Tr with two RDF compressors, HDT and k²-triples. Our experiments show that using RDF-Tr with these compressors improves by up to 2.3 times their original effectiveness, outperforming the most prominent state-of-the-art techniques.

A. Hernández-Illera, M. A. Martínez-Prieto, J. D. Fernández, RDF-TR: Exploiting structural redundancies to boost RDF compression, Information Sciences 508 (2019)