RIQ: Fast processing of SPARQL queries on RDF quadruples

In this paper, we address the problem of fast processing of SPARQL queries on a large RDF dataset, where the RDF statements are quadruples (or quads). Quads can capture provenance or other relevant information about facts. This is especially powerful in modeling knowledge graphs, which are becoming increasingly important on the Web to provide high quality search results to users. We propose a new approach called RIQ that employs a decrease-and-conquer   strategy for fast SPARQL query processing. Rather than indexing the entire RDF dataset, RIQ identifies groups of similar RDF graphs and creates indexes on each group separately. It employs a new vector representation for RDF graphs and locality sensitive hashing to construct the groups efficiently. It constructs a novel filtering index on the groups and compactly represents the index as a combination of Bloom and Counting Bloom Filters. During query processing, RIQ employs a streamlined approach. It constructs a query plan for a SPARQL query (containing one or more graph patterns), searches the filtering index to quickly identify candidate groups that may contain matches for the query, and rewrites the original query to produce an optimized query for each candidate. The optimized queries are then executed using an existing SPARQL processor that supports quads to produce the final results. We conducted a comprehensive evaluation of RIQ using a real and synthetic dataset, each containing about 1.4 billion quads. Our results show that RIQ can outperform its competitors designed to support named graph queries on RDF quads (e.g., Jena TDB and Virtuoso) for a variety of queries.