How to Reuse a Faceted Classification andPut It on the Semantic WebBene Rodriguez-CastroHugh GlaserLes CarrShanghai - 9th ...
Outline• Introduction• Faceted Classification Scheme (FCS)• Dish Detergent FCS Example• FCS Generic Structure• Normalisati...
Introduction (1)• Modeling scenario– There are domain concepts that can berepresented according to multiple classification...
Introduction (2)4colourregiongrapebasetoppinggenderkinshipformbrandscentwinepizzapersondish detergent
Faceted Classification Scheme (FCS)• FCS: a set of mutually exclusive and jointly exhaustivecategories, each made by isola...
Dish Detergent FCS Example (1)6Dish Detergent (Denton, 2003)Facets TermsAgent dishwasher, personForm gel, gelpac, liquid, ...
Dish Detergent FCS Example (2)• Presidents Choice Antibacterial Hand Soap & Dishwashing Liquid– Agent: person– Form: liqui...
FCS Generic Structure• TDC: Target Domain Concept• Faceti: Facet1, Facet2, ..., rest of facets.• FiTermj: Terms or foci or...
Normalisation ODP• Regarded as a best practice and also known as Modularisation, orUntangling (Rector, 2003).• Motivation:...
Normalisation ODP Example (1)• Modeling problem (Egana, 2008)10Egana, 2008. http://www.gong.manchester.ac.uk/odp/html/Norm...
Normalisation ODP Example (2)• Implementation:– Identify the modules– Primitive classes form disjoint homogeneous trees– C...
Normalisation ODP Example (3)12:DefensiveCellrdf:type owl:Class ;rdfs:subClassOf :Cell;owl:equivalentClass [ rdf:type owl:...
Normalisation ODP Example (4)• Result:– The ontology gets untangled– Reasoner maintainssubsumption relations– OWL DL ontol...
Normalisation ODP Generic Structure (1)14:Modulei:MiClassjTDC:SpecificTDCx:TDC:MiClassjowl:topObjectProperty| :performs_fu...
Normalisation ODP Generic Structure (2)15• Elements in the generic structure:– :TDC– :Modulei– :MiClassj– :MiClassjTDC– :S...
FCS Generic Structure• TDC: Target Domain Concept• Faceti: Facet1, Facet2, ..., rest of facets.• FiTermj: Terms or foci or...
Alignment: FCS to Normalisation ODP (1)17• facet in FCSs and module (or semantic axis) in the Normalisation ODP– One persp...
Alignment: FCS to Normalisation ODP (2)InformationScienceOntology ModellingFCS Normalisation ODP FCS in Norm. ODP OWL Impl...
Alignment: FCS to Normalisation ODP (3)19owl:Thing| :Modulei| :MiClassj| :TDC| ( ) :MiClassjTDC| :SpecificTDCxowl:topObjec...
Dish Detergent FCS Example (1)20Dish Detergent (Denton, 2003)Facets TermsAgent dishwasher, personForm gel, gelpac, liquid,...
Dish Detergent FCS Normalisation Example (1)21ComparativeAnalysisowl:Thing| :Faceti| :FiTermj| :TDC| ( ) :FiTermjTDC| :Spe...
Dish Detergent FCS Normalisation Example (2)22owl:Thing| :Agent| :Person| :Dishwasher| :Form| :Gel| :Gelpac| :Liquid| [......
Dish Detergent FCS Example (2)• Presidents Choice Antibacterial Hand Soap & Dishwashing Liquid– Agent: person– Form: liqui...
24:PresidentsPersonLiquidAntibacterialrdf:type owl:Class ;rdfs:subClassOf :DishDetergent ,[ rdf:type owl:Restriction ;owl:...
Dish Detergent FCS Normalisation Example (4)25• FCS Normalised Ontology
Conclusions• Initial set of basic design guidelines to develop an ontology model withinOWL DL that supports the representa...
Future Work• FCS and Normalization ODP: bidirectional alignment?• Multiple FCSs in the same ontology model• Universal Face...
Thanks!Questions and comments:Bene Rodriguez-Castroberoca@gmail.comhttp://purl.org/beroca28
Upcoming SlideShare
Loading in …5
×

How to Reuse a Faceted Classification and Put It on the Semantic Web

418 views

Published on

This presentation illustrate a method to transform a simple faceted classification into an OWL (DL) ontology suitable for deployment on the Semantic Web.
The transformation relies on the alignment between the generic structure of a faceted classification scheme and the Normalization ontology design pattern.

Published in: Technology, Education
0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
418
On SlideShare
0
From Embeds
0
Number of Embeds
1
Actions
Shares
0
Downloads
0
Comments
0
Likes
1
Embeds 0
No embeds

No notes for slide
  • - Ontology EngineeringAims to harness the field with sound development methodologiesAd-hoc practices can lead to unexpected or undesirable results- Correlate “ample room for errors” with “Rector - 2004 - OWL Pizzas Practical Experience of Teaching OWL-DL”.
  • - Consider specifying references.- Welty, C., McGuinness, D.L., Smith, M.K.: OWL web ontology language guide. W3C recommendation, W3C (February 2004), http://www.w3.org/TR/2004/REC-owl-guide-20040210/Krotzsch, M., Patel-Schneider, P.F., Rudolph, S., Hitzler, P., Parsia, B.: OWL 2 web ontology language primer. Technical report, W3C (October 2009), http://www.w3.org/TR/2009/REC-owl2-primer-20091027/- Horridge, M., Drummond, N., Jupp, S., Moulton, G., Stevens, R.: A practical guide to building owl ontologies using the protege-owl plugin and co-ode tools edition 1.2. Technical report, The University Of Manchester (March 2009)
  • -More or less everybody has a notion of what a FCS is.-Nonetheless here are some formal definitions.-In this sense, each facet can be designed separately and it models the domain of discourse from a distinct aspect. -Each facet consists of a terminology, a finite set of terms that exhaust the facet. This set of terms is also referred to as foci.-In our attempt to mitigate this void, we examined a simplified procedure to develop a FCS which contains the conceptualization of various classification criteria (facets) of a specific target domain concept
  • http://www.miskatonic.org/library/facet-web-howto.html
  • -OWL DL compared to the new flavours of OWL 2
  • -OWL DL compared to the new flavours of OWL 2
  • -OWL DL compared to the new flavours of OWL 2
  • -OWL DL compared to the new flavours of OWL 2
  • -OWL DL compared to the new flavours of OWL 2
  • -OWL DL compared to the new flavours of OWL 2
  • -OWL DL compared to the new flavours of OWL 2
  • This approach might not be suitable for converting all possible schemas into a normalised ontology but it is an attempt to provide an initial set of basic design guidelines. These guidelines can be extended hereafter to support more complex cases of FCSs.
  • http://www.miskatonic.org/library/facet-web-howto.html
  • -We can produce an ontology that is a valid instantiation of the norm odp for the fcs example
  • -http://www.gong.manchester.ac.uk/odp/html/Normalisation.htm-this will classify the detergent in question as a: manual, liquid, presidents choice and antibacterial dishwashing detergent.
  • A lack of explicit guidance in the ontology development literature on how to address this recurrent modeling scenario, leaves ample room for ad-hoc practices that can lead to unexpected or undesired results in ontology artifacts
  • - Universal classification. Complications organizing the world’s knowledge.
  • - Universal classification. Complications organizing the world’s knowledge.
  • How to Reuse a Faceted Classification and Put It on the Semantic Web

    1. 1. How to Reuse a Faceted Classification andPut It on the Semantic WebBene Rodriguez-CastroHugh GlaserLes CarrShanghai - 9th ISWC - 8th November 2010
    2. 2. Outline• Introduction• Faceted Classification Scheme (FCS)• Dish Detergent FCS Example• FCS Generic Structure• Normalisation Ontology Design Pattern (ODP)• Normalisation ODP Example• Normalisation ODP Generic Structure• Alignment: FCS to Normalisation ODP• Dish Detergent FCS Normalisation Example• Conclusions• Future Work2
    3. 3. Introduction (1)• Modeling scenario– There are domain concepts that can berepresented according to multiple classificationcriteria– The ontology model requires to represent multipleclassification criteria of the domain concept inquestion– A lack of guidelines in the context of the SemanticWeb leaves ample room for ad-hoc practices3
    4. 4. Introduction (2)4colourregiongrapebasetoppinggenderkinshipformbrandscentwinepizzapersondish detergent
    5. 5. Faceted Classification Scheme (FCS)• FCS: a set of mutually exclusive and jointly exhaustivecategories, each made by isolating one perspective on theitems (a facet), that combine to completely describe all theobjects in question (Denton, 2003).• Facet Analysis: the sorting of terms in a given field ofknowledge into homogeneous, mutually exclusive facets,each derived from the parent universe by a singlecharacteristic of division (Vickery, 1960).• The Principles of Homogeneity and Mutual Exclusivity staterespectively that facets must be homogeneous and mutuallyexclusive, the contents of any two facets cannot overlap, andthat each facet must represent only one characteristic ofdivision of the parent universe (Spiteri, 1998).5
    6. 6. Dish Detergent FCS Example (1)6Dish Detergent (Denton, 2003)Facets TermsAgent dishwasher, personForm gel, gelpac, liquid, powder, tabletBrand Name Cascade, Electrasol, Ivory, No Name, Palmolive,Presidents Choice, SunlightScent green apple, green tea, lavender, lemon, mandarin,ocean breeze, orange blossom, orchard fresh, passionflower, ruby red grapefruit, ylang ylangEffect on Agent aroma therapy (subdivisions: invigorating, relaxing)Special Property antibacterialDenton, W.: How to make a faceted classification and put it on the web (November 2003),http://www.miskatonic.org/library/facet-web-howto.html
    7. 7. Dish Detergent FCS Example (2)• Presidents Choice Antibacterial Hand Soap & Dishwashing Liquid– Agent: person– Form: liquid– Brand Name: Presidents Choice– Scent: (none)– Effect on Agent: (none)– Special Property: antibacterial• Palmolive Aroma Therapy, Lavender and Ylang Ylang– Agent: person– Form: liquid– Brand Name: Palmolive– Scent: lavender, ylang ylang– Effect on Agent: aroma therapy– Special Property: (none) 7
    8. 8. FCS Generic Structure• TDC: Target Domain Concept• Faceti: Facet1, Facet2, ..., rest of facets.• FiTermj: Terms or foci organized by facets:– Facet1: F1Term1, F1Term2, ..., rest of terms in Facet1.– Facet2: F2Term1, F2Term2, ..., rest of terms in Facet2.– ... rest of terms by facet.• Itemx: Set of items (from the TDC) to classify: Item1, Item2,..., rest of items.8
    9. 9. Normalisation ODP• Regarded as a best practice and also known as Modularisation, orUntangling (Rector, 2003).• Motivation: An ontology where a class can have many superclasses(poly-hierarchy).– Subsumption becomes difficult to maintain and error-prone.– Subsumption is implicitly stated.• Aim: to untangle the poly-hierarchy using restrictions to encodesubsumption relationships.– Normalisation allows exactly one unlabelled flavour of is-kind-oflink corresponding to the links declared in the primitive skeleton. Allothers are inferred by the reasoner.– Assertion of multiple inheritance relations among primitiveconcepts are not allowed.9Rector, A.L.: Modularisation of domain ontologies implemented in description logics and related formalisms including owl.In: Proceedings of the 2nd International Conference on Knowledge Capture, K-CAP 2003, pp. 121–128. ACM, New York(2003)
    10. 10. Normalisation ODP Example (1)• Modeling problem (Egana, 2008)10Egana, 2008. http://www.gong.manchester.ac.uk/odp/html/Normalisation.htmlpoly-hierarchyTargetDomainConcept
    11. 11. Normalisation ODP Example (2)• Implementation:– Identify the modules– Primitive classes form disjoint homogeneous trees– Code the subsumption relationships using restrictions– One unlabelled flavour of is-kind-of link11Egana, 2008. http://www.gong.manchester.ac.uk/odp/html/Normalisation.htmlAsserted ontologymodel (Egana, 2008)owl:topObjectProperty| :performs_function
    12. 12. Normalisation ODP Example (3)12:DefensiveCellrdf:type owl:Class ;rdfs:subClassOf :Cell;owl:equivalentClass [ rdf:type owl:Restriction ;owl:onProperty :performs_function;owl:someValuesFrom :Defence ] .:Neutrophilrdf:type owl:Class ;rdfs:subClassOf :AnimalCell,[ rdf:type owl:Restriction ;owl:onProperty :performs_function;owl:someValuesFrom :Defence ] ,[ rdf:type owl:Restriction ;owl:onProperty :performs_function;owl:someValuesFrom :Circulation ] .1-to-1 relationship1-to-many relationship
    13. 13. Normalisation ODP Example (4)• Result:– The ontology gets untangled– Reasoner maintainssubsumption relations– OWL DL ontology model13Egana, 2008. http://www.gong.manchester.ac.uk/odp/html/Normalisation.htmlInferred ontologymodel (Egana, 2008)
    14. 14. Normalisation ODP Generic Structure (1)14:Modulei:MiClassjTDC:SpecificTDCx:TDC:MiClassjowl:topObjectProperty| :performs_function:hasModulei• Elements in the generic structure:– :TDC– :Modulei– :MiClassj– :MiClassjTDC– :SpecificTDCx– :hasModulei
    15. 15. Normalisation ODP Generic Structure (2)15• Elements in the generic structure:– :TDC– :Modulei– :MiClassj– :MiClassjTDC– :SpecificTDCx– :hasModuleiowl:Thing| :Modulei| :MiClassj| :TDC| ( ) :MiClassjTDC| :SpecificTDCxowl:topObjectProperty| :hasModulei( ) denotes a defined class
    16. 16. FCS Generic Structure• TDC: Target Domain Concept• Faceti: Facet1, Facet2, ..., rest of facets.• FiTermj: Terms or foci organized by facets:– Facet1: F1Term1, F1Term2, ..., rest of terms in Facet1.– Facet2: F2Term1, F2Term2, ..., rest of terms in Facet2.– ... rest of terms by facet.• Itemx: Set of items (from the TDC) to classify: Item1, Item2,..., rest of items.16
    17. 17. Alignment: FCS to Normalisation ODP (1)17• facet in FCSs and module (or semantic axis) in the Normalisation ODP– One perspective of the domain being modeled.– A single characteristic of division.– A single criterion of classification.• facets in a FCS to be homogeneous and mutually exclusive.• modules in the Normalisation ODP to be comprised of primitive classesarranged in a structure of disjoint homogeneous class trees.• Both, a facet term and a module subclass, represents the same notion intheir respective conceptual models. A subdivision, a refinement of the facetor module that they complement respectively.• Both, an item in the FCS and a class with multiple parents in theNormalisation ODP, represent the same notion in their respectiveconceptual models. An element that is to be classified.
    18. 18. Alignment: FCS to Normalisation ODP (2)InformationScienceOntology ModellingFCS Normalisation ODP FCS in Norm. ODP OWL ImplementationTDC :TDC owl:Class (primitive)Faceti:Modulei :Faceti owl:Class (primitive):hasModulei :hasFaceti owl:ObjectPropertyFiTermj:MiClassj :FiTermj owl:Class (primitive):MiClassjTDC :FiTermjTDC owl:Class(defined) ( )Itemx :SpecificTDCx owl:Class (primitive)18
    19. 19. Alignment: FCS to Normalisation ODP (3)19owl:Thing| :Modulei| :MiClassj| :TDC| ( ) :MiClassjTDC| :SpecificTDCxowl:topObjectProperty| :hasModuleiTDC: Target Domain ConceptFaceti: Set of facetsFacet1, Facet2, ..., FacetnFiTermj: Terms or foci organized by facetsFacet1: F1Term1, F1Term2, ..., F1TermnFacet2: F2Term1, F2Term2, ..., F2Termn... rest of terms by facetItemx: Set of items to classifyItem1, Item2, ..., ItemnComparativeAnalysisInformationScienceOntology ModellingFCS Norm. ODP FCS in Norm. ODPTDC :TDCFaceti:Modulei :Faceti:hasModulei :hasFacetiFiTermj:MiClassj :FiTermj:MiClassjTDC :FiTermjTDCItemx :SpecificTDCxowl:Thing| :Faceti| :FiTermj| :TDC| ( ) :FiTermjTDC| :SpecificTDCxowl:topObjectProperty| :hasFacetiFCS Generic Structure Normalisation ODP Generic StructureAlignment Guidelines FCS in Normalisation ODPGeneric Structure
    20. 20. Dish Detergent FCS Example (1)20Dish Detergent (Denton, 2003)Facets TermsAgent dishwasher, personForm gel, gelpac, liquid, powder, tabletBrand Name Cascade, Electrasol, Ivory, No Name, Palmolive,Presidents Choice, SunlightScent green apple, green tea, lavender, lemon, mandarin,ocean breeze, orange blossom, orchard fresh, passionflower, ruby red grapefruit, ylang ylangEffect on Agent aroma therapy (subdivisions: invigorating, relaxing)Special Property antibacterialDenton, W.: How to make a faceted classification and put it on the web (November 2003),http://www.miskatonic.org/library/facet-web-howto.html
    21. 21. Dish Detergent FCS Normalisation Example (1)21ComparativeAnalysisowl:Thing| :Faceti| :FiTermj| :TDC| ( ) :FiTermjTDC| :SpecificTDCxowl:topObjectProperty| :hasFacetiDish Detergent FCS FCS in Normalisation ODPGeneric StructureFacet TermsAgent dishwasher, personForm gel, gelpac, liquid, powder, tabletBrand Name Cascade, Electrasol, Ivory, […]Scent green apple, green tea, lavender, […]Effect on Agent aroma therapy (invigorating, relaxing)Special Property Antibacterialowl:Thing| :Agent| :Person| :Dishwasher| :Form| :Gel| :Gelpac| [...]| :BrandName| :Cascade| :Electrasol| [...]| :Scent| :GreenApple| :GreenTea| [...]| :EffectOnAgent| :AromaTherapy| :Invigorating| :Relaxing| :SpecialProperty| :Antibacterialowl:Thing| :DishDetergent (:TDC)| ( ) :ManualDishDetergent| ( ) :DishwasherDishDetergent| ( ) :GelDishDetergent| ( ) :GelpacDishDetergent| ( ) [...]| ( ) :CascaseDishDetergent| ( ) :ElectrasolDishDetergent| ( ) [...]| ( ) :GreenAppleDishDetergent| ( ) :GreenTeaDishDetergent| ( ) [...]| ( ) :AromaTherapyDishDetergent| ( ) :InvigoratingDishDetergent| ( ) :RelaxingDishDetergent| ( ) :AntibacterialDishDetergent| :PresidentsPersonLiquidAntibacterial| :PalmoliveAromaTherapyLavenderYlangYlang| :SpecificDishDetergent3| (... rest of :SpecificDishDetergentx)Dish Detergent FCSNormalised Ontology
    22. 22. Dish Detergent FCS Normalisation Example (2)22owl:Thing| :Agent| :Person| :Dishwasher| :Form| :Gel| :Gelpac| :Liquid| [...]| :BrandName| :Cascade| :Electrasol| :Ivory| [...]| :Scent| :GreenApple| :GreenTea| :Lavender| [...]| :EffectOnAgent| :AromaTherapy| :Invigorating| :Relaxing| :SpecialProperty| :Antibacterialowl:Thing| :DishDetergent (:TDC)| ( ) :ManualDishDetergent| ( ) :DishwasherDishDetergent| ( ) :GelDishDetergent| ( ) :GelpacDishDetergent| ( ) :LiquidDishDetergent| ( ) [...]| ( ) :CascaseDishDetergent| ( ) :ElectrasolDishDetergent| ( ) :IvoryDishDetergent| ( ) [...]| ( ) :GreenAppleDishDetergent| ( ) :GreenTeaDishDetergent| ( ) :LavenderDishDetergent| ( ) [...]| ( ) :AromaTherapyDishDetergent| ( ) :InvigoratingDishDetergent| ( ) :RelaxingDishDetergent| ( ) :AntibacterialDishDetergent| :PresidentsPersonLiquidAntibacterial| :PalmoliveAromaTherapyLavenderYlangYlang| :SpecificDishDetergent3| (... rest of :SpecificDishDetergentx):hasAgent• FCS Normalised Ontologyhttp://purl.org/net/project/enakting/ontology/detergent_fcs_norm
    23. 23. Dish Detergent FCS Example (2)• Presidents Choice Antibacterial Hand Soap & Dishwashing Liquid– Agent: person– Form: liquid– Brand Name: Presidents Choice– Scent: (none)– Effect on Agent: (none)– Special Property: antibacterial• Palmolive Aroma Therapy, Lavender and Ylang Ylang– Agent: person– Form: liquid– Brand Name: Palmolive– Scent: lavender, ylang ylang– Effect on Agent: aroma therapy– Special Property: (none) 23
    24. 24. 24:PresidentsPersonLiquidAntibacterialrdf:type owl:Class ;rdfs:subClassOf :DishDetergent ,[ rdf:type owl:Restriction ;owl:onProperty :hasAgent;owl:someValuesFrom :Person ] ,[ rdf:type owl:Restriction ;owl:onProperty :hasForm;owl:someValuesFrom :Liquid ] ,[ rdf:type owl:Restriction ;owl:onProperty :hasBrandName;owl:someValuesFrom :PresidentsChoice ] ,[ rdf:type owl:Restriction ;owl:onProperty :hasSpecialProperty;owl:someValuesFrom :Antibacterial ] ,Dish Detergent FCS Normalisation Example (3)
    25. 25. Dish Detergent FCS Normalisation Example (4)25• FCS Normalised Ontology
    26. 26. Conclusions• Initial set of basic design guidelines to develop an ontology model withinOWL DL that supports the representation of multiple classificationcriteria of a specific domain concept.• A series of mappings between the elements of a generic FCS and theNormalization ODP have been identified that allow us to convert a givenFCS into an OWL DL ontology model following a consistent andsystematic approach.• The guidelines presented in this first effort consider explicitly theconceptualization of existing classification criteria in the context of ontologymodeling for the Semantic Web and provide a partial solution to theproblem scenario described.• An existing FCS example in the domain of “Dishwashing Detergent” isused to illustrate the main steps of our conversion procedure26
    27. 27. Future Work• FCS and Normalization ODP: bidirectional alignment?• Multiple FCSs in the same ontology model• Universal Faceted Classifications• Revisiting the Normalization ODP using OWL 2 meta-modelling (punning)• :MiClassj and :SpecificTDCx - owl:Class orowl:NamedIndividual27
    28. 28. Thanks!Questions and comments:Bene Rodriguez-Castroberoca@gmail.comhttp://purl.org/beroca28

    ×