\n
The Description \nLogic Benchmark Suite
\n\n\n
\nThis DL Benchmark Suite is based on benchmark problems for the DL98 \nSystem Comparison with addtional ABox benchmarks and extended \napplication knowledge bases. In addition, the TABLEAUX/TANCS'99 \nbenchmark problems are integrated. Many people have contributed to this \ncollection (see the acknowledgements).\n\n
\nWith contributions from \nVolker \nHaarslev, \nIan Horrocks,\nRalf \nMöller and\n\nPeter Patel-Schneider.\n\n\n
\n\n
\nSee also the \nDL'99 Systems Comparison at the \n\n1999 International Workshop on Description Logics - DL'99\n\n
\nIf you have additional problem sets to be included into the benchmark \nsuite, please send a message.\n\n
\n\n
Download the DL Benchmark Suite
\n\n\n\nCurrently available DL systems that can be used to run \nthe DL Benchmark Suite
\n\n- \n
- \nDLP: Included with source code in \nSML for Unix \nand Windows \n
- \nFaCT: Version 1.11 included with source code in \nANSI Common Lisp for Macintosh, Unix and Windows \n
- RACE:\nVersion 1.1 included for
\n- \n\t
- Macintosh Common Lisp 4.2/4.3 (free evaluation copy) \n\t
- Allegro Common Lisp 4.3 Solaris \n\t
- Allegro Common Lisp 5.0 Linux (free \n\tevaluation copy) \n\t
- Allegro Common Lisp 5.0 Solaris (free evaluation \n\tcopy) \n\t
- Allegro Common Lisp 5.0.1 Windows (free evaluation \n\tcopy) \n\t
- For teaching purposes see also RACE Online \n
\n
For instructions on how to run these systems on the benchmarks see the \nfile readme.html. A complete list of description logic systems \nis available from the description \nlogic home page.\n\n
\n\n\n
New benchmarks compared to the DL98 System Comparison
\n\n- \n
- New application knowledge bases (e.g. TBoxes with GCIs). \n
- Additional and more demanding ABox tests. \n
- Problems of (un)satisfiability in modal logic K or, in equivalent terms, \n of concept satisfiability in the description\n logic ALC (with only one role) and empty TBoxes. \n
- Problems that are still in PSPACE but require the use of global axioms\n when encoded in modal logic. \n
More demanding ABox benchmarks
\n\n\nThe initial DL'98 ABox benchmarks are given as a set of inference problems \nconsisting of a TBox, an ABox and a set of instance checking queries. \nHowever, for optimized systems such as RACE, these DL'98 ABox benchmarks \nare far too simple, i.e. the execution time is almost not measurable. Even \nfor the hardest problems the runtimes are below 3 seconds (realization \ninclusive). Therefore, we have compiled a new set of benchmarks for ALC \nABoxes based on more complex TBoxes as before (based on the TBoxes of the \nDL'98 System Comparison). Each benchmark now consists of a set of \nsubproblems which, in turn, consist of an ALC TBox, an ALC ABox and a set \nof instance checking problems which should evaluate to true. We generated a \nset of nine new ABox benchmark problem sets, each with up to 21 problems. \nThe current versions of FaCT and DLP do not support ABox reasoning. Thus, \nonly RACE can be used to run the ABox tests.\n\n
More demanding application knowledge bases
\n\nWith respect to the DL'98 System Comparison new application knowledge bases \nhave been added. The DL'98 versions did not use role hierachies and domain \nand range restrictions for roles. The new application knowledge bases can \nbe run with RACE. FaCT 1.11 does not support number restrictions. DLP \ncannot handle GCIs.\n\nEvaluation Results
\n\n- \n
- Evalutation results for different systems have been published \n \n here. \n
- For the new DL system RACE the evaluation results (Macintosh version) can\nbe found here. \n
\n\n
Acknowledgments
\n\nWe would like to thank \nFabio Massaci for the \npermission to integrate the TANCS'99 benchmark problems.\n\n\n\n\n