Revisiting SPARQL's Semantics for Blank Nodes

This page contains complementary material describing the experiments, settings, and results obtained with that settings. Also, this material can be useful to repeat the experiments with different settings in order to produce comparable results.

  • Wikidata queries. This file includes all Wikidata's example queries on April 07, 2018.
  • Queries. All queries used in the experiment. This compressed folder includes three directories: low for low level joins, tpch for queries based in the TPC-H benchmark, and wikidata for Wikidata's example queries (including only 3 Wikidata queries that produce false positives and the fixed version of the queries).
  • Generator of low level joins dataset. This compressed folder has the scripts to build the low level joins dataset.
  • Generator of the TPC-H based dataset. This compressed folder has the scripts to build the TPCH-based dataset. This scripts must be used in combination with the official TPC-H generator, that because its license restrictions is not published here.
  • Data loader. This compressed file has the scripts to load datasets in Fuseki and Virtuoso.
  • Run benchmarks. This compressed file has the scripts to run the benchmarks.
  • Instructions to run the experiments. This is a document containing the instructions to run the low level joins and TPC-H based experiments.
  • Result data from TPC-H based experiment in Virtuoso and Fuseki. These files include the summaries of outputs and elapsed times for each TPCH query. The name of each file describes the query and dataset used. For instance, file ds-sf030-br001-q4.3-summary.json corresponds to a dataset with scale factor 3, blank rate 1 and query template q4.3. The file contains three records, one for a different set of attributes: the params used to compose the query, the number of results, if the query execution finished before the timeout (i.e., before 300 seconds), and the elapsed time (in seconds).
  • Average times by scale. This compressed folder contains avg times for low level joins with respect to the size of the database.
  • Analysis of Wikidata queries. Document analyzing queries that are a candidate for having false positives.