The paper “DBpedia SPARQL Benchmark – Performance Assessment with Real Queries on Real Data” won the best paper award at the world’s most prestigious Semantic Web conference, ISWC. LIMES, one of the technologies developed by InFAI during the course of the SCMS project, played a central role in the generation of this benchmark.The basic observation behind the paper was that most of the current SPARQL benchmarks were extracted from data that reflect relational schemas (small number of properties and classes) and not necessarily native RDF data. The aim of the paper was to generate a benchmark that addresses this drawback by mining the query log of one of the best reknown knowledge bases in the world, DBpedia . The resulting benchmark consists of 25 queries that can be used to assess the characteristics of triple stores when they are confronted with native RDF data. One of the main challenges during the mining process was to compute the similarity of queries in a time-efficient manner without loss of recall. The LIMES framework was used for this purpose and reduced the runtime of whole algorithm to less than 17% of the original runtime. The new version of LIMES (described here) can achieve the same computation orders of magnitude faster. Also remember to check out the new GUI.
Link on,
Axel
