SweetProlog: A System to Integrate Ontologies and Rules

Loredana Laera, Valentina Tamma and Trevor Bench-CaponUniversity of LiverpoolLiverpool L69 7ZF, UK

{lori, valli, tbc@csc.liv.ac.uk}

Abstract

In this poster, we present our prototype system torepresent semantic derivation rules to support rea-soning on the Semantic Web. The prototype, calledSweetProlog, is able to reason over the instancesof an ontology within rule definitions using a Pro-log engine and enables the integration of ontologiesand rules.

1 IntroductionRules represent an important issue for the Semantic Web andconstitute the next language level over the ontology languageson the Semantic Web, as envisioned by Tim-Berners Lee[Berners-Leeet al., 2001]. Rules give the ability to draw in-ferences, to express constraints and to describe some aspectsof applications, not expressible in ontology languages. In thiscontext, several rule languages and standards have been pro-posed, such as RuleML[Boleyet al., 2001] and SWRL[Hor-rocks et al., 2004], a RuleML-oriented extension of OWL.However, the SWRL proposal provides a syntactical and se-mantical interoperation between ontologies and rules but notinferential, due to its undecidability. The crucial point is toprovide some reasoning support and to providesemantic, syn-tactic and inferential interoperation between ontologies andrules. In other words, we need a way to enable the rules touse the vocabulary of an ontology and to reason in a semanti-cally consistent way to draw inferences.Here we propose our prototype for reasoning, both with on-tologies and rules, that uses a Logic Programming Language,such as Prolog, to offer efficient automatic reasoning. Thesystem, called SweetProlog, shows how derivation rules maybe used to enhance the content of a ontology and allow thedynamic inclusion of derived facts that are not captured bythe ontological taxonomy alone.

2 Features of SweetProlog2.1 A “meta-ontology ”for RuleMLIn order to define the rules we have used a OWL ontologyof Courteous Logic Programs (CLP) RuleML rule syntax[Grosof et al., 2002], a sub-language of the Rule MarkupLanguage. Intuitively, the resulting language, called OWL-RuleML, can be viewed as a”meta-ontology”of the RuleML

syntax. OWLRuleML allows to specify derivation rules andpreserves the semantics of all constraints in RuleML. An in-teresting aspect of OWLRuleML is the possibility to repre-sent non-monotonicity via Courteous Logic Programs. CLPs[Grosof, 1997] extend the expressivity of Ordinary Logic Pro-grams (OLP), and they are tractably compilable to OLP by aCourteous Compiler. CLPs provide a method to resolve con-flicts between rules using partially prioritized information toguarantee a consistent and unique set of conclusions (answer-set).

2.2 Rules on top of ontologiesIn order to integrate rules with OWL ontologies, we followthe approach proposed[Grosof, 2003] to build rules on top ofontologies. Principally, the OWLRuleML rules can be linkedto descriptions in OWL. The names of predicates in the OWL-RuleML rules are URI’s that link to classes and propertiesin an OWL ontology. The OWL ontology that is referencedforms a background theory for the rulebase. Formally, theKnowledge Representation contained within the intersectionof LP rules on top of DL ontologies corresponds to Descrip-tion Logics Programs (DLP)[Grosofet al., 2003]. DLP is afragment of OWL corresponding to Horn clauses, i.e. to thelogic programming fragment of OWL. DLP provide a signif-icant degree of expressiveness, substantially greater than theRDFS fragment of Description Logics.

2.3 Translation of OWL in PrologChoosing the DLP fragment imposes some restrictions onOWL DL in order to guarantee that all axioms stated are inter-pretable with Prolog. In general, the sub-set of OWL that canbe evaluated with Prolog includes the whole of the OWL frag-ment of RDF Schema, the equality of classes, properties andindividuals of OWL, all property characteristics, cardinalityconstraints with minimum cardinality 0 and maximum cardi-nality, range restrictions on properties, conjunction of classesand construction of classes by enumeration. The result is nec-essarily less expressive than description logics. However, asshown in[Volz, 2004] very few available ontologies use con-structs outside the DLP language fragment.

3 SweetProlog ArchitectureSweetProlog enables reasoning over OWL ontologies byrules via a translation of OWL subsets into simple Prolog

Figure 1: Rules on top of Ontologies

predicates, which a SWI-Prolog engine can understand andprocess. SweetProlog is implemented in Java and makesuse of different components (see Figure 2): JPL1, a Java in-terface to Prolog, for providing abridge between Java andSWI-Prolog; a set of inference rules to translate OWL intoProlog; IBM Courteous Compiler2 to resolve conflicts be-tween rules using partially prioritized information and Jena23

to parse OWL ontologies. SweetProlog extracts RDF triples

Figure 2: Architecture overview

by parsing OWL ontologies and OWLRuleML rules. TheRDF triples that represent OWL concepts and instances aretranslated into Prolog predicates via a set of inference rulesto translate OWL into Prolog. The RDF triples that representthe rules are translated into Courteous Logic Program rules.Such rules are transformed by a Courteous Compiler into Pro-log rules. Finally, the translated predicates are then fed intothe working memory of a SWI-Prolog engine to infer new

1http://www.swi-prolog.org/packages/jpl2http://alphaworks.ibm.com/tech/commonrules3http://jena.sourceforge.net

knowledge. SweetProlog also includes an editor that permitsthe interactive editing of OWLRuleML rules.

4 ConclusionsIn this poster we have presented a prototype application thatattempts to provide semantic, syntactic and inferential inter-operation between OWL ontologies and RuleML rules. Themain strength of the approach is that is provides not onlyreasoning support for OWLRuleML rules, via a OWLmeta-ontologyfor RuleML syntax, but also reasoning support forOWL ontologies combined with OWLRuleML rules. Thisapproach develops Grosof and Horrocks’s idea of specifyingRuleML rules on top of OWL ontologies, suggesting a map-ping of a Description Logic subset (OWL) into Logic Pro-grams (Prolog). The intersection of DL with LP, called De-scription Logic Programs, covers RDF Schema and a signifi-cant fragment of OWL.

5 AcknowledgementThe research presented in this paper is funded by PIPS (FP6-IST No 507019)

References[Berners-Leeet al., 2001] T. Berners-Lee, J. Hendler, and

O. Lassila. The Semantic Web.Scientific American.284(5):3443, May, 2001.

[Boleyet al., 2001] H. Boley, S. Tabet, and G. Wagner. De-sign rationale of RuleML: A markup language for seman-tic web rules. InInternational Semantic Web WorkingSymposium (SWWS), 2001.

[Grosof, 1997] B. Grosof. Courteous logic programs: Prior-itized conflict handling for rules. Technical report,IBMResearch Report RC 20836, Dec. 30 1997, revised fromMay 8 1997.

[Grosof, 2003] B. Grosof. SweetDeal: Representing AgentContracts with Exceptions using XML Rules, Ontologies,and Process Descriptions. InProceedings of the twelfthinternational conference on World Wide Web, Budapest,Hungary, 2003.

[Grosofet al., 2002] B. Grosof, M. D. Gandhe andT. W. Finin. SweetJess: Translating DAMLRuleMLto JESS. InProceedings of the International Workshopon Rule Markup Languages for Business Rules on theSemantic Web, Sardinia, Italy,2002.

[Grosofet al., 2003] Benjamin N. Grosof, Ian Horrocks,Raphael Volz, and Stefan Decker. Description LogicPrograms: Combining Logic Programs with DescriptionLogic. In Proc. of the Twelfth International World WideWeb Conference (WWW 2003), pages 48–57. ACM, 2003.

[Horrockset al., 2004] I. Horrocks, P. F. Patel-Schneider,H. Boley, S. Tabet, B. Grosof and M. Dean.SWRL: A Se-mantic Web Rule Language Combining OWL and RuleML.W3C Member Submission, 21 May 2004.

[Volz, 2004] R. Volz. Web Ontology Reasoning with LogicDatabases.PhD thesis, AIFB, University of Karlsruhe,2004.