## A Comparison of Reasoning Techniques for Querying Large Description Logic ABoxes (2006)

### Cached

### Download Links

- [web.comlab.ox.ac.uk]
- [www.cs.man.ac.uk]
- [www.cs.ox.ac.uk]
- DBLP

### Other Repositories/Bibliography

Citations: | 54 - 10 self |

### BibTeX

@INPROCEEDINGS{Motik06acomparison,

author = {Boris Motik and Ulrike Sattler},

title = {A Comparison of Reasoning Techniques for Querying Large Description Logic ABoxes},

booktitle = {},

year = {2006},

pages = {227--241},

publisher = {Springer}

}

### OpenURL

### Abstract

Abstract. Many modern applications of description logics (DLs) require answering queries over large data quantities, structured according to relatively simple ontologies. For such applications, we conjectured that reusing ideas of deductive databases might improve scalability of DL systems. Hence, in our previous work, we developed an algorithm for reducing a DL knowledge base to a disjunctive datalog program. To test our conjecture, we implemented our algorithm in a new DL reasoner KAON2, which we describe in this paper. Furthermore, we created a comprehensive test suite and used it to conduct a performance evaluation. Our results show that, on knowledge bases with large ABoxes but with simple TBoxes, our technique indeed shows good performance; in contrast, on knowledge bases with large and complex TBoxes, existing techniques still perform better. This allowed us to gain important insights into strengths and weaknesses of both approaches. 1

### Citations

413 | RACER system description
- Haarslev, Möller
(Show Context)
Citation Context ... preserving the set of relevant consequences. Thissalgorithm is quite different from tableau algorithms [1, Chapter 2] and their optimizations [9], used in state-of-the-art DL reasoners such as RACER =-=[8]-=-, FaCT++ [22], or Pellet [17]. We conjectured that our algorithm will scale well to knowledge bases with large ABoxes and simple TBoxes. In particular, we expected great benefits from techniques such ... |

308 | Otter 3.0 reference manual and guide
- McCune
- 1994
(Show Context)
Citation Context ...-xmlsyntax/sis implemented in the Theorem Prover subcomponent of the Reasoning Engine. Although there are several efficient theorem provers for first-order logic (e.g., Vampire [19], E [20], or Otter =-=[14]-=-), we decided to implement our own theorem prover, due to the following reasons. First, we are unaware of an existing implementation of basic superposition. Second, existing systems usually do not com... |

208 | Reducing OWL Entailment to Description Logic Satisfiability
- Horrocks, Patel-Schneider
- 2003
(Show Context)
Citation Context ... computer science. They have long been used in information integration [1, Chapter 16], and they provide a logical foundation for OWL—a standardized language for ontology modeling in the Semantic Web =-=[10]-=-. A DL knowledge base is typically partitioned into a terminological (or schema) part, called a TBox, and an assertional (or data) part, called an ABox. Whereas some applications rely on reasoning ove... |

181 |
The Design and Implementation of Vampire
- Riazanov, Voronkov
- 2002
(Show Context)
Citation Context ...ttp://www.w3.org/TR/owl-xmlsyntax/sis implemented in the Theorem Prover subcomponent of the Reasoning Engine. Although there are several efficient theorem provers for first-order logic (e.g., Vampire =-=[19]-=-, E [20], or Otter [14]), we decided to implement our own theorem prover, due to the following reasons. First, we are unaware of an existing implementation of basic superposition. Second, existing sys... |

169 | Practical reasoning for very expressive description logics
- Horrocks, Sattler, et al.
(Show Context)
Citation Context ...easoning techniques for expressive DLs, allowing the users to choose the one that best suits the needs of their application. 2 Preliminaries We now present the syntax and the semantics of the DL SHIQ =-=[11]-=-—the formalism underlying KAON2. Given a set of role names NR, a SHIQ role is either some R ∈ NR or an inverse role R − for R ∈ NR. A SHIQ RBox KB R is a 1 http://kaon2.semanticweb.org/ 2 http://kaon2... |

136 | E: A Brainiac Theorem Prover
- Schulz
(Show Context)
Citation Context ...w.w3.org/TR/owl-xmlsyntax/sis implemented in the Theorem Prover subcomponent of the Reasoning Engine. Although there are several efficient theorem provers for first-order logic (e.g., Vampire [19], E =-=[20]-=-, or Otter [14]), we decided to implement our own theorem prover, due to the following reasons. First, we are unaware of an existing implementation of basic superposition. Second, existing systems usu... |

131 | Reducing SHIQ− description logic to disjunctive datalog programs
- Motik, Sattler, et al.
- 2004
(Show Context)
Citation Context ...that have proven themselves in practice. Motivated by the prospect of applying these techniques to query answering in description logics, in our previous work we described a novel reasoning algorithm =-=[12]-=- that reduces a SHIQ knowledge base KB to a disjunctive datalog program DD(KB) while preserving the set of relevant consequences. Thissalgorithm is quite different from tableau algorithms [1, Chapter ... |

121 |
Optimising Tableaux Decision Procedures for Description Logics
- Horrocks
- 1997
(Show Context)
Citation Context ...dge base KB to a disjunctive datalog program DD(KB) while preserving the set of relevant consequences. Thissalgorithm is quite different from tableau algorithms [1, Chapter 2] and their optimizations =-=[9]-=-, used in state-of-the-art DL reasoners such as RACER [8], FaCT++ [22], or Pellet [17]. We conjectured that our algorithm will scale well to knowledge bases with large ABoxes and simple TBoxes. In par... |

104 | A deductive system for nonmonotonic reasoning
- Eiter, Leone, et al.
- 1997
(Show Context)
Citation Context ...ct of reasoning in disjunctive datalog. To compute the models, disjunctive datalog engines usually ground the disjunctive program. Although this process has been optimized using intelligent grounding =-=[6]-=-, grounding can be very expensive on large data sets. In contrast,sthe models of our programs are of no interest. To avoid grounding, we answer queries using hyperresolution with answer literals. Four... |

103 | Data complexity of reasoning in very expressive description logics
- Hustadt, Motik, et al.
- 2005
(Show Context)
Citation Context ...s with large ABoxes and simple TBoxes. In particular, we expected great benefits from techniques such as magic sets [4] or join-order optimizations. Furthermore, we identified a Horn fragment of SHIQ =-=[13]-=-, for which our algorithm exhibits polynomial data complexity (that is, the complexity measured in the size of the ABox, assuming the TBox is fixed in size). To test our conjecture, we implemented the... |

84 | An evaluation of knowledge base systems for large OWL datasets
- Guo, Pan, et al.
- 2004
(Show Context)
Citation Context ...2(x, y, z) ≡ Man(x),isCreditCardOf (y, x),Gold(y),livesIn(x, z),Region(z) LUBM 9 is a benchmark developed at the Lehigh University for testing performance of ontology management and reasoning systems =-=[7]-=-. The ontology describes organizational structure of universities and it is relatively simple: it does not use disjunctions or number restrictions, but it does use existential quantifiers, so our redu... |

76 | Reasoning in Description Logics using Resolution and Deductive Databases
- Motik
- 2006
(Show Context)
Citation Context ... paper we outline the design of the system and overview the employed techniques. Due to lack of space, we are unable to present all optimizations in full detail; for more information, please refer to =-=[15]-=-. Providing an objective account of the performance of our approach proved to be difficult because there are no widely recognized benchmarks for query answering. To fill this gap, we created a benchma... |

73 | Theorem Proving with Ordering and Equality Constrained Clauses
- Nieuwenhuis, Rubio
- 1995
(Show Context)
Citation Context ... effects of all mentioned inference steps, without expanding the R-successors explicitly. Computing all relevant “macro” rules is performed by saturating the TBox of KB using basic superposition (BS) =-=[2,16]-=- (a clausal refutation calculus), which 3 http://jena.sourceforge.net/ 4 http://www.w3.org/TR/owl-semantics/ 5 http://www.w3.org/TR/owl-xmlsyntax/sis implemented in the Theorem Prover subcomponent of ... |

67 | Basic Paramodulation
- Bachmair, Ganzinger, et al.
- 1995
(Show Context)
Citation Context ... effects of all mentioned inference steps, without expanding the R-successors explicitly. Computing all relevant “macro” rules is performed by saturating the TBox of KB using basic superposition (BS) =-=[2,16]-=- (a clausal refutation calculus), which 3 http://jena.sourceforge.net/ 4 http://www.w3.org/TR/owl-semantics/ 5 http://www.w3.org/TR/owl-xmlsyntax/sis implemented in the Theorem Prover subcomponent of ... |

64 | Declarative problem-solving using the DLV system
- Eiter, Faber, et al.
- 2000
(Show Context)
Citation Context ...g Engine subcomponent of the Reasoning Engine is used for answering queries in the disjunctive datalog program obtained by the reduction. Although several disjunctive datalog engines exist (e.g., DLV =-=[5]-=-), we decided to implement our own engine, due to the following reasons. First, existing engines do not come with a comprehensive API, which makes their integration into other systems difficult. Secon... |

63 | Cooking the semantic web with the owl api
- Bechhofer, Lord, et al.
- 2003
(Show Context)
Citation Context ...on services, such as adding and retrieving axioms. The API fully supports OWL and the Semantic Web Rule Language (SWRL) at the syntactic level. Several similar APIs already exist, such as the OWL API =-=[3]-=- or Jena. 3 However, to obtain an efficient system, we needed complete control over the internals of the API, and could thus not reuse an existing implementation. Ontologies can be saved in files, usi... |

58 |
Web Ontology Reasoning with Logic Databases
- Volz
- 2004
(Show Context)
Citation Context ...t seem unfair. However, we did not yet consider caching for KAON2; furthermore, materialized views were extensively studied in deductive databases, and were successfully applied to ontology reasoning =-=[23]-=-. Also, KAON2 does not perform a separate ABox consistency test because ABox inconsistency is discovered automatically during query evaluation; we consider this to be an advantage of our approach. Due... |

34 | Goals for concept representation in the galen project
- Rector, Nowlan, et al.
- 1993
(Show Context)
Citation Context ...127 10086 wine 8 20007 19926 wine 9 39767 39606 wine 10 79287 78966 dolce 203 27 42 2 253 253 522 0 0 galen 3237 699 0 133 0 0 287 0 0 GALEN 12 is a medical terminology developed in the GALEN project =-=[18]-=-. It has a very large and complex TBox and no ABox, and has traditionally been used as a benchmark for terminological reasoning. Table 2 shows the number of axioms for each ontology. 5 Performance Eva... |

27 | Enhancing the magic-set method for disjunctive datalog programs
- Cumbo, Faber, et al.
(Show Context)
Citation Context ... or Pellet [17]. We conjectured that our algorithm will scale well to knowledge bases with large ABoxes and simple TBoxes. In particular, we expected great benefits from techniques such as magic sets =-=[4]-=- or join-order optimizations. Furthermore, we identified a Horn fragment of SHIQ [13], for which our algorithm exhibits polynomial data complexity (that is, the complexity measured in the size of the ... |

15 |
Ordering Heuristics for Description Logic Reasoning
- Tsarkov, Horrocks
(Show Context)
Citation Context ...the set of relevant consequences. Thissalgorithm is quite different from tableau algorithms [1, Chapter 2] and their optimizations [9], used in state-of-the-art DL reasoners such as RACER [8], FaCT++ =-=[22]-=-, or Pellet [17]. We conjectured that our algorithm will scale well to knowledge bases with large ABoxes and simple TBoxes. In particular, we expected great benefits from techniques such as magic sets... |

11 | Simple and efficient clause subsumption with feature vector indexing
- Schulz
- 2004
(Show Context)
Citation Context ...ary and binary literals containing shallow terms, for which unification can be implemented in constant time. Furthermore, clauses can be efficiently indexed using a variant of feature vector indexing =-=[21]-=-. The Ontology Clausification subcomponent of the Reasoning Engine is responsible for translating the TBox of a SHIQ knowledge base KB into a set of first-order clauses. As our experiments confirm, it... |

8 |
Pellet: An OWL-DL Reasoner. Poster
- Parsia, Sirin
- 2004
(Show Context)
Citation Context ...ant consequences. Thissalgorithm is quite different from tableau algorithms [1, Chapter 2] and their optimizations [9], used in state-of-the-art DL reasoners such as RACER [8], FaCT++ [22], or Pellet =-=[17]-=-. We conjectured that our algorithm will scale well to knowledge bases with large ABoxes and simple TBoxes. In particular, we expected great benefits from techniques such as magic sets [4] or join-ord... |