ontology alignment representation
DESCRIPTION
An ontology alignment representation frameworkTRANSCRIPT
Ontology Alignment Representation
Francois Scharffe
Semantic Technology Institute (STI)University of Innsbruck, Austria
April 14, 2008
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 1 / 34
Outline
IntroductionOntology mediationExample
Actual limitations of the approach
Three levels of knowledge abstractionSPARQL for instance transformationThe Alignment Format and Ontology
Correspondence PatternsPatternsPattern templatePattern libraryDatabase mapping to ontologies patterns
Conclusion
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 2 / 34
Introduction Ontology mediation
Situation
I Many ontologies overlap
I Need to exchange data between applications/services/agents
I Ontology mediation is the set of techniques making this dataexchange possible
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 3 / 34
Introduction Ontology mediation
Situation
I Many ontologies overlap
I Need to exchange data between applications/services/agents
I Ontology mediation is the set of techniques making this dataexchange possible
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 3 / 34
Introduction Ontology mediation
Situation
I Many ontologies overlap
I Need to exchange data between applications/services/agents
I Ontology mediation is the set of techniques making this dataexchange possible
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 3 / 34
Introduction Ontology mediation
Two phases ontology mediation
Two phases can be distiguished in ontology mediation:
I Constructing an alignment (matching, graphical tool).
I Processing the alignment for a mediation task (data translation,ontology merging, query rewriting)
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 4 / 34
Introduction Ontology mediation
Two phases ontology mediation
Two phases can be distiguished in ontology mediation:
I Constructing an alignment (matching, graphical tool).
I Processing the alignment for a mediation task (data translation,ontology merging, query rewriting)
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 4 / 34
Introduction Ontology mediation
Ontology Alignment
Definition (Alignment, correspondence)
Given two ontologies o and o ′, an alignment between o and o ′ is a set ofcorrespondences (i.e., 4-uples): 〈e, e ′, r , n〉 with
I e ∈ o and e ′ ∈ o ′ being the two matched entities,
I r being a relationship holding between e and e ′, and
I n expressing the level of confidence [0..1] in this correspondence.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 5 / 34
Introduction Example
Example 1: wines from Bordeaux
A BordeauxWine is a wine produced in the Bordeaux region.
ClassCorrespondence
wine:BordeauxWine vin:Vin
vin:terroir
geo:LocationAttribute value restriction
Attribute value restriction pattern
http://.../Château-Margaux
http://.../Château-Marguaux
vin:terroir
geo:Bordelais
Figure: Bordeaux wines
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 6 / 34
Actual limitations of the approach
Problem
I while many matching algorithms, many graphical tools are out there,there is not something such as a common format that would allow toexchange alignments they output.
I while the correspondence such as the one in the precedent exampleoccurs frequently, no algorithm is able to automatically detect it.
I We cope with these issues in the following way
I We provide an alignment representation formalism allowingexchange of alignments through their lifecycle
I We formalize aspects of recurring correspondences and extractpatterns that will provide support for constructing alignments.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 7 / 34
Actual limitations of the approach
Problem
I while many matching algorithms, many graphical tools are out there,there is not something such as a common format that would allow toexchange alignments they output.
I while the correspondence such as the one in the precedent exampleoccurs frequently, no algorithm is able to automatically detect it.
I We cope with these issues in the following way
I We provide an alignment representation formalism allowingexchange of alignments through their lifecycle
I We formalize aspects of recurring correspondences and extractpatterns that will provide support for constructing alignments.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 7 / 34
Actual limitations of the approach
Problem
I while many matching algorithms, many graphical tools are out there,there is not something such as a common format that would allow toexchange alignments they output.
I while the correspondence such as the one in the precedent exampleoccurs frequently, no algorithm is able to automatically detect it.
I We cope with these issues in the following way
I We provide an alignment representation formalism allowingexchange of alignments through their lifecycle
I We formalize aspects of recurring correspondences and extractpatterns that will provide support for constructing alignments.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 7 / 34
Actual limitations of the approach
Problem
I while many matching algorithms, many graphical tools are out there,there is not something such as a common format that would allow toexchange alignments they output.
I while the correspondence such as the one in the precedent exampleoccurs frequently, no algorithm is able to automatically detect it.
I We cope with these issues in the following way
I We provide an alignment representation formalism allowingexchange of alignments through their lifecycle
I We formalize aspects of recurring correspondences and extractpatterns that will provide support for constructing alignments.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 7 / 34
Actual limitations of the approach
Problem
I while many matching algorithms, many graphical tools are out there,there is not something such as a common format that would allow toexchange alignments they output.
I while the correspondence such as the one in the precedent exampleoccurs frequently, no algorithm is able to automatically detect it.
I We cope with these issues in the following way
I We provide an alignment representation formalism allowingexchange of alignments through their lifecycle
I We formalize aspects of recurring correspondences and extractpatterns that will provide support for constructing alignments.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 7 / 34
Three levels of knowledge abstraction
I three levels of abstraction for ontology alignment representation
I at the ground level: mediation rules or grounded correspondences(expressed in OWL, COWL, SWRL, WSML variants, SPARQL++,etc.).
I at the intermediate level: alignment (expressed in a particular rule orobject model of a tool/algorithm)
I at the top level of abstraction: correspondence patterns (newconcept)
We implement this framework on the three levels.
I at the ground level: a SPARQL extension for data transformation
I at the intermediate level: the Alignment format and the AlignmentOntology
I at the top level: a library of recurring correspondence patterns
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 8 / 34
Three levels of knowledge abstraction
I three levels of abstraction for ontology alignment representation
I at the ground level: mediation rules or grounded correspondences(expressed in OWL, COWL, SWRL, WSML variants, SPARQL++,etc.).
I at the intermediate level: alignment (expressed in a particular rule orobject model of a tool/algorithm)
I at the top level of abstraction: correspondence patterns (newconcept)
We implement this framework on the three levels.
I at the ground level: a SPARQL extension for data transformation
I at the intermediate level: the Alignment format and the AlignmentOntology
I at the top level: a library of recurring correspondence patterns
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 8 / 34
Three levels of knowledge abstraction
I three levels of abstraction for ontology alignment representation
I at the ground level: mediation rules or grounded correspondences(expressed in OWL, COWL, SWRL, WSML variants, SPARQL++,etc.).
I at the intermediate level: alignment (expressed in a particular rule orobject model of a tool/algorithm)
I at the top level of abstraction: correspondence patterns (newconcept)
We implement this framework on the three levels.
I at the ground level: a SPARQL extension for data transformation
I at the intermediate level: the Alignment format and the AlignmentOntology
I at the top level: a library of recurring correspondence patterns
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 8 / 34
Three levels of knowledge abstraction
I three levels of abstraction for ontology alignment representation
I at the ground level: mediation rules or grounded correspondences(expressed in OWL, COWL, SWRL, WSML variants, SPARQL++,etc.).
I at the intermediate level: alignment (expressed in a particular rule orobject model of a tool/algorithm)
I at the top level of abstraction: correspondence patterns (newconcept)
We implement this framework on the three levels.
I at the ground level: a SPARQL extension for data transformation
I at the intermediate level: the Alignment format and the AlignmentOntology
I at the top level: a library of recurring correspondence patterns
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 8 / 34
Three levels of knowledge abstraction
I three levels of abstraction for ontology alignment representation
I at the ground level: mediation rules or grounded correspondences(expressed in OWL, COWL, SWRL, WSML variants, SPARQL++,etc.).
I at the intermediate level: alignment (expressed in a particular rule orobject model of a tool/algorithm)
I at the top level of abstraction: correspondence patterns (newconcept)
We implement this framework on the three levels.
I at the ground level: a SPARQL extension for data transformation
I at the intermediate level: the Alignment format and the AlignmentOntology
I at the top level: a library of recurring correspondence patterns
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 8 / 34
Three levels of knowledge abstraction
I three levels of abstraction for ontology alignment representation
I at the ground level: mediation rules or grounded correspondences(expressed in OWL, COWL, SWRL, WSML variants, SPARQL++,etc.).
I at the intermediate level: alignment (expressed in a particular rule orobject model of a tool/algorithm)
I at the top level of abstraction: correspondence patterns (newconcept)
We implement this framework on the three levels.
I at the ground level: a SPARQL extension for data transformation
I at the intermediate level: the Alignment format and the AlignmentOntology
I at the top level: a library of recurring correspondence patterns
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 8 / 34
Three levels of knowledge abstraction
I three levels of abstraction for ontology alignment representation
I at the ground level: mediation rules or grounded correspondences(expressed in OWL, COWL, SWRL, WSML variants, SPARQL++,etc.).
I at the intermediate level: alignment (expressed in a particular rule orobject model of a tool/algorithm)
I at the top level of abstraction: correspondence patterns (newconcept)
We implement this framework on the three levels.
I at the ground level: a SPARQL extension for data transformation
I at the intermediate level: the Alignment format and the AlignmentOntology
I at the top level: a library of recurring correspondence patterns
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 8 / 34
Three levels of knowledge abstraction SPARQL for instance transformation
SPARQL++ for instance transformation
We propose to use SPARQL as a convenient data translation language forRDF.
I it allows to query data
I its CONSTRUCT statement allows to translate graphs fragments
I It’s a W3C recommendation and is already supported by numeroustools
I However, features of SPARQL are not mature enough to expresscomplex mappings: we thus propose an extension:SPARQL++
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 9 / 34
Three levels of knowledge abstraction SPARQL for instance transformation
SPARQL++ for instance transformation
We propose to use SPARQL as a convenient data translation language forRDF.
I it allows to query data
I its CONSTRUCT statement allows to translate graphs fragments
I It’s a W3C recommendation and is already supported by numeroustools
I However, features of SPARQL are not mature enough to expresscomplex mappings: we thus propose an extension:SPARQL++
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 9 / 34
Three levels of knowledge abstraction SPARQL for instance transformation
SPARQL++ for instance transformation
We propose to use SPARQL as a convenient data translation language forRDF.
I it allows to query data
I its CONSTRUCT statement allows to translate graphs fragments
I It’s a W3C recommendation and is already supported by numeroustools
I However, features of SPARQL are not mature enough to expresscomplex mappings: we thus propose an extension:SPARQL++
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 9 / 34
Three levels of knowledge abstraction SPARQL for instance transformation
SPARQL++ for instance transformation
We propose to use SPARQL as a convenient data translation language forRDF.
I it allows to query data
I its CONSTRUCT statement allows to translate graphs fragments
I It’s a W3C recommendation and is already supported by numeroustools
I However, features of SPARQL are not mature enough to expresscomplex mappings: we thus propose an extension:SPARQL++
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 9 / 34
Three levels of knowledge abstraction SPARQL for instance transformation
Simple example (Translate from VCard to FOAF)
vCard
VCard:FN
FOAF
foaf:name
⊆
CONSTRUCT { ?X foaf:name ?Y }WHERE { ?X VCard:FN ?Y }
Easy! Supported by standard SPARQL
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 10 / 34
Three levels of knowledge abstraction SPARQL for instance transformation
Simple example (Translate from VCard to FOAF)
vCard
VCard:FN
FOAF
foaf:name
⊆
CONSTRUCT { ?X foaf:name ?Y }WHERE { ?X VCard:FN ?Y }
Easy! Supported by standard SPARQL
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 10 / 34
Three levels of knowledge abstraction SPARQL for instance transformation
Built-in functions and value generation
vCard FOAF
foaf:name
VCard:Family
VCard:Given
⊆
CONSTRUCT { ?X foaf:name fn:concat(?N," ",?F }WHERE { ?X VCard:Given ?N. ?X VCard:Family ?F }
This is not possible in standard SPARQL
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 11 / 34
Three levels of knowledge abstraction SPARQL for instance transformation
Built-in functions and value generation
vCard FOAF
foaf:name
VCard:Family
VCard:Given
⊆
CONSTRUCT { ?X foaf:name fn:concat(?N," ",?F }WHERE { ?X VCard:Given ?N. ?X VCard:Family ?F }
This is not possible in standard SPARQL
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 11 / 34
Three levels of knowledge abstraction SPARQL for instance transformation
Example: Aggregates
Translate from DOAP to RDF Open Source Software Vocabulary
CONSTRUCT { ?P os:latestReleaseMAX(?V : ?P doap:release ?R. ?R doap:revision ?V) }
WHERE { ?P rdf:type doap:Project . }
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 12 / 34
Three levels of knowledge abstraction SPARQL for instance transformation
Example: Aggregates
Translate from DOAP to RDF Open Source Software Vocabulary
CONSTRUCT { ?P os:latestReleaseMAX(?V : ?P doap:release ?R. ?R doap:revision ?V) }
WHERE { ?P rdf:type doap:Project . }
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 12 / 34
Three levels of knowledge abstraction SPARQL for instance transformation
Example: regular path expressions (pSPARQL)
o1
o1:parentOf
o2
o2:ancestorOf
⊆+
CONSTRUCTs with regular path expressions can cater for that, this ispossible in pSPARQL:
CONSTRUCT { ?X o2:ancestor ?Y }WHERE { ?X o1:parentOf+ ?Y . }
pSPARQL offers even more: full regular expressions over paths conditionsover paths, etc.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 13 / 34
Three levels of knowledge abstraction SPARQL for instance transformation
Example: regular path expressions (pSPARQL)
o1
o1:parentOf
o2
o2:ancestorOf
⊆+
CONSTRUCTs with regular path expressions can cater for that, this ispossible in pSPARQL:
CONSTRUCT { ?X o2:ancestor ?Y }WHERE { ?X o1:parentOf+ ?Y . }
pSPARQL offers even more: full regular expressions over paths conditionsover paths, etc.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 13 / 34
Three levels of knowledge abstraction SPARQL for instance transformation
Example: regular path expressions (pSPARQL)
o1
o1:parentOf
o2
o2:ancestorOf
⊆+
CONSTRUCTs with regular path expressions can cater for that, this ispossible in pSPARQL:
CONSTRUCT { ?X o2:ancestor ?Y }WHERE { ?X o1:parentOf+ ?Y . }
pSPARQL offers even more: full regular expressions over paths conditionsover paths, etc.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 13 / 34
Three levels of knowledge abstraction SPARQL for instance transformation
Example: regular path expressions (pSPARQL)
o1
o1:parentOf
o2
o2:ancestorOf
⊆+
CONSTRUCTs with regular path expressions can cater for that, this ispossible in pSPARQL:
CONSTRUCT { ?X o2:ancestor ?Y }WHERE { ?X o1:parentOf+ ?Y . }
pSPARQL offers even more: full regular expressions over paths conditionsover paths, etc.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 13 / 34
Three levels of knowledge abstraction SPARQL for instance transformation
I SPARQL++ was published at OTM 2007. The paper contains detailsof the extension, semantics, and implementation details.
I pSPARQL integration in our alignment framework was published atOnAv 2008.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 14 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
Motivation and requirements
I considering the alignment as a first-class citizen
I being independent from the ontology formalism
I being able to represent correspondences between ontological entities
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 15 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
Motivation and requirements
I considering the alignment as a first-class citizen
I being independent from the ontology formalism
I being able to represent correspondences between ontological entities
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 15 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
Motivation and requirements
I considering the alignment as a first-class citizen
I being independent from the ontology formalism
I being able to represent correspondences between ontological entities
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 15 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
The Alignment Ontology
I Models the ontology alignment domain
I Classes, Attributes, Relations and Instances for the entities
I One to one, many to many, using set operators on the entities
I Heterogeneous correspondences
I Conditions can restrict entities scope
I Transformations of attributes values
I Paths
I Model theoretic based semantics
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 16 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
The Alignment Ontology
I Models the ontology alignment domain
I Classes, Attributes, Relations and Instances for the entities
I One to one, many to many, using set operators on the entities
I Heterogeneous correspondences
I Conditions can restrict entities scope
I Transformations of attributes values
I Paths
I Model theoretic based semantics
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 16 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
The Alignment Ontology
I Models the ontology alignment domain
I Classes, Attributes, Relations and Instances for the entities
I One to one, many to many, using set operators on the entities
I Heterogeneous correspondences
I Conditions can restrict entities scope
I Transformations of attributes values
I Paths
I Model theoretic based semantics
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 16 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
The Alignment Ontology
I Models the ontology alignment domain
I Classes, Attributes, Relations and Instances for the entities
I One to one, many to many, using set operators on the entities
I Heterogeneous correspondences
I Conditions can restrict entities scope
I Transformations of attributes values
I Paths
I Model theoretic based semantics
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 16 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
The Alignment Ontology
I Models the ontology alignment domain
I Classes, Attributes, Relations and Instances for the entities
I One to one, many to many, using set operators on the entities
I Heterogeneous correspondences
I Conditions can restrict entities scope
I Transformations of attributes values
I Paths
I Model theoretic based semantics
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 16 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
The Alignment Ontology
I Models the ontology alignment domain
I Classes, Attributes, Relations and Instances for the entities
I One to one, many to many, using set operators on the entities
I Heterogeneous correspondences
I Conditions can restrict entities scope
I Transformations of attributes values
I Paths
I Model theoretic based semantics
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 16 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
The Alignment Ontology
I Models the ontology alignment domain
I Classes, Attributes, Relations and Instances for the entities
I One to one, many to many, using set operators on the entities
I Heterogeneous correspondences
I Conditions can restrict entities scope
I Transformations of attributes values
I Paths
I Model theoretic based semantics
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 16 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
The Alignment Ontology
I Models the ontology alignment domain
I Classes, Attributes, Relations and Instances for the entities
I One to one, many to many, using set operators on the entities
I Heterogeneous correspondences
I Conditions can restrict entities scope
I Transformations of attributes values
I Paths
I Model theoretic based semantics
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 16 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
The Alignment Format
Example serialization of Cell instances (RDF/XML):FOAF persons based in Innsbruck have a particular VCard phone extension:
<Cell><entity1><Class rdf:about="&foaf;Person"><attributeValueCondition><Restriction><onProperty rdf:resource="&foaf;based_near"/><comparator rdf:datatype="&xsd;string">xsd:equals</comparator><value rdf:datatype="&xsd;string">Fortaleza</value>
</Restriction></attributeValueCondition></Class>
</entity1>
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 17 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
The alignment format as an exchange format (2)
<entity2><Class rdf:about="&v;VCard"><attributeValueCondition><Restriction><onProperty rdf:resource="&v;workTel"/><comparator rdf:datatype="&xsd;string">xsd:startsWith</comparator><value rdf:datatype="&xsd;string">+0043512</value></Restriction>
</attributeValueCondition></Class>
</entity2><measure RDF:datatype=’&BSD;float’>1.0</measure><relation>equivalence</relation></Cell>
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 18 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
Alignments semantics
Based on [Zimmermann 2006] and largely compliant with descriptionlogics.
Syntax level o o ′
Local semantics level D D ′I I ′
Global semantics level Dε ε′
A
I
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 19 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
Alignments semantics
Based on [Zimmermann 2006] and largely compliant with descriptionlogics.
Syntax level o o ′
Local semantics level D D ′I I ′
Global semantics level Dε ε′
A
I
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 19 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
Alignments semantics
Based on [Zimmermann 2006] and largely compliant with descriptionlogics.
Syntax level o o ′
Local semantics level D D ′I I ′
Global semantics level Dε ε′
A
I
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 19 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
Alignments semantics
Based on [Zimmermann 2006] and largely compliant with descriptionlogics.
Syntax level o o ′
Local semantics level D D ′I I ′
Global semantics level Dε ε′
A
I
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 19 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
Entity semantics
The semantics of expressions are interpreted within D through:
∀x ∈ QL(o), ε o I (x) ∈ D
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 20 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
Expression semantics
The expression semantics is typically the semantics of a description logic:
I (c) = ε o I (c)
I (C t C ′) = I (C ) ∪ I (C ′)
I (C u C ′) = I (C ) ∩ I (C ′)
I (¬C ) = D − I (C )
I (∃K ) = {x ∈ D|∃y ∈ D; 〈x , y〉 ∈ I (K )}
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 21 / 34
Three levels of knowledge abstraction The Alignment Format and Ontology
Alignment processing
The combined alignment API (INRIA) and Mapping API allow to parsealignments and ground them for execution.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 22 / 34
Correspondence Patterns Patterns
Patterns in the literature
I Capture recurring knowledge
I Facilitate the design task by using known structures
I Facilitate semi-automatic design
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 23 / 34
Correspondence Patterns Patterns
Patterns in the literature
I Capture recurring knowledge
I Facilitate the design task by using known structures
I Facilitate semi-automatic design
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 23 / 34
Correspondence Patterns Patterns
Patterns in the literature
I Capture recurring knowledge
I Facilitate the design task by using known structures
I Facilitate semi-automatic design
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 23 / 34
Correspondence Patterns Pattern template
Pattern template elements
I Name: give a meaningful name for the pattern, corresponds to afragment of the URI associated with the pattern.
I Alias: alternative name
I Problem: A statement describing the intent, goals of the patterns,which entities are involved in the correspondence.
I Context: refers to the context of usage for the pattern (specificdomain, specific application)
I Solution: describe the result of using the pattern.
I Example: example alignment representing this pattern
I Related Patterns: relationships between this pattern and others
I Known uses: URIs of existing alignments using this patterns
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 24 / 34
Correspondence Patterns Pattern template
Pattern template elements
I Name: give a meaningful name for the pattern, corresponds to afragment of the URI associated with the pattern.
I Alias: alternative name
I Problem: A statement describing the intent, goals of the patterns,which entities are involved in the correspondence.
I Context: refers to the context of usage for the pattern (specificdomain, specific application)
I Solution: describe the result of using the pattern.
I Example: example alignment representing this pattern
I Related Patterns: relationships between this pattern and others
I Known uses: URIs of existing alignments using this patterns
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 24 / 34
Correspondence Patterns Pattern template
Pattern template elements
I Name: give a meaningful name for the pattern, corresponds to afragment of the URI associated with the pattern.
I Alias: alternative name
I Problem: A statement describing the intent, goals of the patterns,which entities are involved in the correspondence.
I Context: refers to the context of usage for the pattern (specificdomain, specific application)
I Solution: describe the result of using the pattern.
I Example: example alignment representing this pattern
I Related Patterns: relationships between this pattern and others
I Known uses: URIs of existing alignments using this patterns
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 24 / 34
Correspondence Patterns Pattern template
Pattern template elements
I Name: give a meaningful name for the pattern, corresponds to afragment of the URI associated with the pattern.
I Alias: alternative name
I Problem: A statement describing the intent, goals of the patterns,which entities are involved in the correspondence.
I Context: refers to the context of usage for the pattern (specificdomain, specific application)
I Solution: describe the result of using the pattern.
I Example: example alignment representing this pattern
I Related Patterns: relationships between this pattern and others
I Known uses: URIs of existing alignments using this patterns
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 24 / 34
Correspondence Patterns Pattern template
Pattern template elements
I Name: give a meaningful name for the pattern, corresponds to afragment of the URI associated with the pattern.
I Alias: alternative name
I Problem: A statement describing the intent, goals of the patterns,which entities are involved in the correspondence.
I Context: refers to the context of usage for the pattern (specificdomain, specific application)
I Solution: describe the result of using the pattern.
I Example: example alignment representing this pattern
I Related Patterns: relationships between this pattern and others
I Known uses: URIs of existing alignments using this patterns
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 24 / 34
Correspondence Patterns Pattern template
Pattern template elements
I Name: give a meaningful name for the pattern, corresponds to afragment of the URI associated with the pattern.
I Alias: alternative name
I Problem: A statement describing the intent, goals of the patterns,which entities are involved in the correspondence.
I Context: refers to the context of usage for the pattern (specificdomain, specific application)
I Solution: describe the result of using the pattern.
I Example: example alignment representing this pattern
I Related Patterns: relationships between this pattern and others
I Known uses: URIs of existing alignments using this patterns
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 24 / 34
Correspondence Patterns Pattern template
Pattern template elements
I Name: give a meaningful name for the pattern, corresponds to afragment of the URI associated with the pattern.
I Alias: alternative name
I Problem: A statement describing the intent, goals of the patterns,which entities are involved in the correspondence.
I Context: refers to the context of usage for the pattern (specificdomain, specific application)
I Solution: describe the result of using the pattern.
I Example: example alignment representing this pattern
I Related Patterns: relationships between this pattern and others
I Known uses: URIs of existing alignments using this patterns
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 24 / 34
Correspondence Patterns Pattern template
Pattern template elements
I Name: give a meaningful name for the pattern, corresponds to afragment of the URI associated with the pattern.
I Alias: alternative name
I Problem: A statement describing the intent, goals of the patterns,which entities are involved in the correspondence.
I Context: refers to the context of usage for the pattern (specificdomain, specific application)
I Solution: describe the result of using the pattern.
I Example: example alignment representing this pattern
I Related Patterns: relationships between this pattern and others
I Known uses: URIs of existing alignments using this patterns
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 24 / 34
Correspondence Patterns Pattern template
Example pattern: Aggregation pattern
os:Software
doap:realease
os:LatestVersion
doap:Version
Hoary Hedgehog
Warty Warthog
...Hardy Heron
doap:realease
"4.10"
"8.04"...
"5.04"
doap:Project
Attribute Value aggregation (MAX)
"8.04"
Attribute Correspondence
ClassCorrespondence
valuesdoap:revision
Aggregation Pattern
Stringhttp://www.ubuntu.org/
os:LatestVersion
doap:revision
doap:revision
doap:revision...
http://www.ubuntu.org/
Figure: Aggregation Pattern
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 25 / 34
Correspondence Patterns Pattern template
Example pattern: RCA-A pattern
RCA-A concatPattern
vc:VCard
vc:family-name
vc:Nfoaf:name
foaf:Personvc:name
value
Attribute RelationCorrespondence
Class AttributeCorrespondence
Value TransformationConcatenation
vc:given-name
vc:additional-name
http://.../francois
vc:name
_:bn01
"Scharffe"
"François"
"Vincent Alfred"
http://.../francois
foaf:name
"François Vincent Alfred Scharffe"
Figure: Relation Class Attribute to Attribute Pattern
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 26 / 34
Correspondence Patterns Pattern library
Pattern library
I Implemented as an ontology extending the Alignment Ontology withspecific patterns properties.
I Sets of placeholders entities to be replaced during patternsinstantiation.
I Around 40 patterns available.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 27 / 34
Correspondence Patterns Pattern library
Pattern library
I Implemented as an ontology extending the Alignment Ontology withspecific patterns properties.
I Sets of placeholders entities to be replaced during patternsinstantiation.
I Around 40 patterns available.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 27 / 34
Correspondence Patterns Pattern library
Pattern library
I Implemented as an ontology extending the Alignment Ontology withspecific patterns properties.
I Sets of placeholders entities to be replaced during patternsinstantiation.
I Around 40 patterns available.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 27 / 34
Correspondence Patterns Pattern library
Pattern library(2)
Figure: Overview of the patterns library
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 28 / 34
Correspondence Patterns Database mapping to ontologies patterns
Database mapping to ontologies
Direct Mapping A table in the database directly corresponds to a conceptin the ontology. There is a one-to-one correspondencebetween records in the table and instances of the concept.This pattern is modeled on the basis of the Class equivalencecorrespondence pattern.
Join/Union A set of database tables corresponds to a concept in theontology when they are joined. There is a one-to-onecorrespondence between join records of the joined tables andinstances of an ontology concept. This pattern is modeledon the Class Union pattern.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 29 / 34
Correspondence Patterns Database mapping to ontologies patterns
Database mapping to ontologies (2)
Projection A subset of the columns of a database table are needed tomap a concept in the ontology. This pattern is modeled onthe basis of the Class By Attribute Type pattern, where thescope of the class (the table) is restricted to the specificattributes (columns).
Selection A subset of the rows of a database table map a concept inthe ontology. This pattern is modeled on the basis of theClass By Attribute Value, rrestrictingthe scope of the class(the table) to only those instances (the table rows) having agiven value for the specified aattributes(columns).
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 30 / 34
Correspondence Patterns Database mapping to ontologies patterns
Database mapping to ontologies (3)
Value transformation A column in an table corresponds to an attribute inthe ontology after transformation of its value using atransformation function. This pattern is modeled on thebasis of the Attribute Transformation pattern.
Combinations Combinations of the aforementioned patterns are alsopossible.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 31 / 34
Conclusion
Conclusion
This thesisI Introduces a new three layered framework reflecting the ontology
mediation landscape:
1. Mediation rules with the introduction of a SPARQL extension forinstance transformation.
2. Ontology Alignment with the introduction of a alignmentrepresentation formalism: language and ontology.
3. Correspondence patterns introduced, formalized, reference patternslibrary.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 32 / 34
Conclusion
Conclusion
This thesisI Introduces a new three layered framework reflecting the ontology
mediation landscape:
1. Mediation rules with the introduction of a SPARQL extension forinstance transformation.
2. Ontology Alignment with the introduction of a alignmentrepresentation formalism: language and ontology.
3. Correspondence patterns introduced, formalized, reference patternslibrary.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 32 / 34
Conclusion
Conclusion
This thesisI Introduces a new three layered framework reflecting the ontology
mediation landscape:
1. Mediation rules with the introduction of a SPARQL extension forinstance transformation.
2. Ontology Alignment with the introduction of a alignmentrepresentation formalism: language and ontology.
3. Correspondence patterns introduced, formalized, reference patternslibrary.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 32 / 34
Conclusion
Conclusion
This thesisI Introduces a new three layered framework reflecting the ontology
mediation landscape:
1. Mediation rules with the introduction of a SPARQL extension forinstance transformation.
2. Ontology Alignment with the introduction of a alignmentrepresentation formalism: language and ontology.
3. Correspondence patterns introduced, formalized, reference patternslibrary.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 32 / 34
Conclusion
Ongoing and Future works
I Write the thesis !
I Groundings to SPARQL++
I Extend the pattern library with domain and application specificpatterns.
I Study how patterns can assist matching algorithms to discovercomplex matches.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 33 / 34
Conclusion
Ongoing and Future works
I Write the thesis !
I Groundings to SPARQL++
I Extend the pattern library with domain and application specificpatterns.
I Study how patterns can assist matching algorithms to discovercomplex matches.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 33 / 34
Conclusion
Ongoing and Future works
I Write the thesis !
I Groundings to SPARQL++
I Extend the pattern library with domain and application specificpatterns.
I Study how patterns can assist matching algorithms to discovercomplex matches.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 33 / 34
Conclusion
Ongoing and Future works
I Write the thesis !
I Groundings to SPARQL++
I Extend the pattern library with domain and application specificpatterns.
I Study how patterns can assist matching algorithms to discovercomplex matches.
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 33 / 34
Conclusion
Thank you for your attention !
Questions ?
Francois Scharffe (STI Innsbruck) Ontology Alignment Representation April 14, 2008 34 / 34