algebraic graph transformations for merging ontologies
DESCRIPTION
The conception of an ontology is a complex task influenced by numerous factors like the point of view of the authors or the level of details. Consequently, several ontologies have been developed to model identical or related domains leading to partially overlapping representations. This divergence of conceptualization requires the study of ontologies merging in order to create a common repository of knowledge and integrate various sources of information. In this work, we propose a formal approach for merging ontologies using typed graph grammars. This method relies on the algebraic approach to graph transformations, SPO (Simple PushOut) which allows a formal representation and ensures the consistence of the results. Furthermore, a new ontologies merging algorithm called GROM (Graph Rewriting for Ontology Merging) is presented.TRANSCRIPT
Context Typed Graph Grammars Approach Implementation Conclusion References
Algebraic Graph Transformations for MergingOntologies
Mariem MAHFOUDH, Laurent THIRY, Germain FORESTIERand Michel HASSENFORDER
{mariem.mahfoudh, laurent.thiry, germain.forestier, michel.hassenforder}@uha.fr
UHA–MIPS, EA 2332 /ENSISA, 12 rue des Freres Lumiere68093 Mulhouse Cedex, France
International Conference on Model & Data Engineering (MEDI 2014)Larnaca, Cyprus, 25 September 2014
M.Mahfoudh et al. Merging Ontologies with SPO 1/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Outline
1 Context
2 Typed Graph Grammars
3 Approach
4 Implementation
5 Conclusion
M.Mahfoudh et al. Merging Ontologies with SPO 2/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Outline
1 Context
2 Typed Graph Grammars
3 Approach
4 Implementation
5 Conclusion
M.Mahfoudh et al. Merging Ontologies with SPO 3/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Context
European project CCAlps (Creative Companies in Alpine space)
Aims:
I Help and facilitate the collaboration between creativecompanies in Alpine Space
I Link several companies and partners to organize events
Idea
I Use ontologies to model data and build a consensus betweenthe various partners
Web site: www.ccalps.eu
M.Mahfoudh et al. Merging Ontologies with SPO 4/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Ontologies
I Ontologies are an explicit specification of a conceptualizationof a domain [Gruber, 1993]
M.Mahfoudh et al. Merging Ontologies with SPO 5/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Ontologies
I Ontologies are an explicit specification of a conceptualizationof a domain [Gruber, 1993]
M.Mahfoudh et al. Merging Ontologies with SPO 5/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Ontologies
I Ontologies are an explicit specification of a conceptualizationof a domain [Gruber, 1993]
⇒ Require concepts, methods and tools to merge ontologies
M.Mahfoudh et al. Merging Ontologies with SPO 5/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Merging ontologies
I Merging ontologies is defined as ”Creating a new ontologyfrom two or more existing ontologies with overlapping parts,which can be either virtual or physical” [Klein, 2001]
Similarity search
Merging ontologies:Create a global ontology
Symmetric
Asymmetric
O1 O2
GO
Merging ontologies process
Similarity search
Merging ontologies:Create a global ontology
Symmetric
Asymmetric
O1 O2
GO
Merging ontologies process
⇒ Define an asymmetric approach for merging ontologies using TypedGraph Grammars to reduce the conflicts in the Global Ontology
M.Mahfoudh et al. Merging Ontologies with SPO 6/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Merging ontologies
I Merging ontologies is defined as ”Creating a new ontologyfrom two or more existing ontologies with overlapping parts,which can be either virtual or physical” [Klein, 2001]
Similarity search
Merging ontologies:Create a global ontology
Symmetric
Asymmetric
O1 O2
GO
Merging ontologies process
⇒ Define an asymmetric approach for merging ontologies using TypedGraph Grammars to reduce the conflicts in the Global Ontology
M.Mahfoudh et al. Merging Ontologies with SPO 6/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Outline
1 Context
2 Typed Graph Grammars
3 Approach
4 Implementation
5 Conclusion
M.Mahfoudh et al. Merging Ontologies with SPO 7/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Typed Graph Grammars
I Formalism for the representation and management of graphs
I Let TGG = (G ,GT ,P) be a typed graph grammar where:
B G = (V ,E ) is a host graphB GT = (VT ,ET ) is a type graph
PACMan game with TGG, [Ehrig and Padberg, 2004].
M.Mahfoudh et al. Merging Ontologies with SPO 8/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Typed Graph Grammars
I Formalism for the representation and management of graphs
I Let TGG = (G ,GT ,P) be a typed graph grammar where:
B G = (V ,E ) is a host graphB GT = (VT ,ET ) is a type graph
PACMan game with TGG, [Ehrig and Padberg, 2004].
M.Mahfoudh et al. Merging Ontologies with SPO 8/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Typed Graph Grammars
TGG = (GT ,G ,P)
I P = (LHS ,RHS) is a set of production rules:
B LHS (Left Hand Side) represents the precondition of therewriting rule and defines the structure that has to be found inG .
B RHS (Right Hand Side) represents the postcondition of therewriting rule and must replaces LHS in G .
⇒ NAC (Negative Application Condition) specify the graphthat should not occur when matching a rule.
More details: [Rozenberg, 1999]M.Mahfoudh et al. Merging Ontologies with SPO 9/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Simple PushOut (SPO)
I Applying a rewriting rule to an initial graph (G ) with the SPOmethod consists in:
1 Finding LHS in G by defining a morphism m : LHS → G2 Deleting m(LHS)−m(LHS ∩ RHS) from G3 Adding m(RHS)−m(LHS ∩ RHS) to G to get a new version
G ′
m
Rewriting rule with PACMan game [Ehrig and Padberg, 2004].
M.Mahfoudh et al. Merging Ontologies with SPO 10/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Outline
1 Context
2 Typed Graph Grammars
3 Approach
4 Implementation
5 Conclusion
M.Mahfoudh et al. Merging Ontologies with SPO 11/25
Context Typed Graph Grammars Approach Implementation Conclusion References
A graph transformation model for ontology changes
TGG = (G ,GT ,P)
I G describes the ontology
I GT represents the ontologies meta-model
Type Graph for representing ontologies.
M.Mahfoudh et al. Merging Ontologies with SPO 12/25
Context Typed Graph Grammars Approach Implementation Conclusion References
A graph transformation model for ontology changes
TGG = (G ,GT ,P)
I P represents the ontologies changes
G
LHS RHS
domain
name="Location"
Class
name="hasLocation"
ObjectProperty
name="Company"
1:Class
name="Entreprise"
1:Class
NAC
name="Entreprise"
Class
RenameClass
m
name="Company"Class
range
G'
domain
name="Location"
Class
name="hasLocation"
ObjectProperty
name="Entreprise"Class
range
M. Mahfoudh, G. Forestier, L. Thiry, and M. Hassenforder, ”Consistent ontologiesevolution using graph grammars”, International Conference on Knowledge Science,Engineering and Management, Dalian (Chine), pages 64-75, Springer, 2013.
M.Mahfoudh et al. Merging Ontologies with SPO 13/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Merging ontologies with SPO
Approach
I Asymmetric approach for merging ontologies
I Approach based on algebraic graph transformations
O1 O2
Step1: Similarity search
Step 2: Merge
Step 3: Adaptation
CO O2
O1 GO GO'
LHSCH RHSCH
SPO
SPO*
Approach of merging ontologies with algebraic graph transformations.
M.Mahfoudh et al. Merging Ontologies with SPO 14/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Step 1: similarity search
Similarity search
I Define the correspondences between the ontologies O1 andO2:
1 Common nodes (CN)2 Equivalents nodes (EN)3 Nodes sharing a semantic relation (SN)4 Nodes sharing a subsumption relation (IsaN)
M.Mahfoudh et al. Merging Ontologies with SPO 15/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Step 1: similarity search
EOCCAlps ontology extract COCCAlps ontology extract
Similarity search example
I CN = {”Tag”}; EN = {(”hasTag”, ”has Tag”)}I SN = {(”Emplacement”, ”Location”)}I IsaN = {(”Company”, ”Participant”)}
M.Mahfoudh et al. Merging Ontologies with SPO 16/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Step 2: merging ontologies
EN{O1} EN{O2}
O1 O'1
CO O2
O'1 GO
(LHS) (RHS)
(G) (G')
(LHS) (RHS)
(G) (G')
O'1 O2
CO
Sub-step1 Sub-step2 Sub-step3
Merging ontologies
1 Minimize the differences between the two ontologies.⇒ ∀N ∈ {EN} doO ′
1 ← SPO RenameEntity (O1, EN{O1}, EN{O2} )
2 Create the common ontology (CO)
3 Create the global ontology (GO)⇒ GO ← SPO MergeGraph (O ′
1, CO, O2)
M.Mahfoudh et al. Merging Ontologies with SPO 17/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Step 3: global ontology adaptation
Global ontology adaptation
I Enrich the global ontology by adding the semantic andsubsumption relations
LHS
1:Class
name="C1"
2:Class
name="C2"
RHS
2:Class
name="C2"
1:Class
name="C1"
subClassOf subClassOf
NAC5
1:Class
name="C1"
disjointwith
Class
2:Class
name="C2"
Class
subClassOf subClassOf
subClassOf
NAC1
1:Class
name="C2"
2:Class
name="C1"
NAC2
1:Class
name="C2"
2:Class
name="C1"
disjointWith
NAC3
1:Class
name="C2"
2:Class
name="C1"
NAC4
2:Class
name="C2"
subClassOf
1:Class
name="C1"
Class
subClassOf
Rewriting rule for the AddSubClassChange.
M.Mahfoudh et al. Merging Ontologies with SPO 18/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Step 3: global ontology adaptation
Result of merging the ontologies O1&O2
Global ontology adaptation example
I SN = {(”Emplacement”, ”Location”)}I IsaN = {(”Company”, ”Participant”)}
M.Mahfoudh et al. Merging Ontologies with SPO 19/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Outline
1 Context
2 Typed Graph Grammars
3 Approach
4 Implementation
5 Conclusion
M.Mahfoudh et al. Merging Ontologies with SPO 20/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Implementation
I Implementation of two java prototypes:
B OntologiesMapping is based on WordNet ontology andLevenshtein distance
B GROM (Graph Rewriting for Ontology Merging) is based onAGG (Algebraic Graph Grammar) API
OntologiesMapping
GROM
O1 (.owl)
O2 (.owl)
MappingFile(.xml)
GO (.ggx)
GGXToOWL
OWLToGGX
O1 (.ggx)
O2 (.ggx)
GO (.owl)
Web site: http://mariem-mahfoudh.info/medi2014M.Mahfoudh et al. Merging Ontologies with SPO 21/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Outline
1 Context
2 Typed Graph Grammars
3 Approach
4 Implementation
5 Conclusion
M.Mahfoudh et al. Merging Ontologies with SPO 22/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Conclusion
Summary
I Propose an asymmetric approach for merging ontologies usingtyped graph grammar and algebraic graph transformations
I Implement the approach and propose GROM tool
Future work
I Study large ontologies
I Study the conflicts in details
M.Mahfoudh et al. Merging Ontologies with SPO 23/25
Context Typed Graph Grammars Approach Implementation Conclusion References
References
Hartmut Ehrig and Julia Padberg. Graph grammars and petri net transformations. InLectures on Concurrency and Petri Nets, pages 496–536. Springer, 2004.
Thomas R. Gruber. A translation approach to portable ontology specifications.Knowledge acquisition, 5(2):199–220, 1993.
Michel Klein. Combining and relating ontologies: an analysis of problems andsolutions. In IJCAI-2001, Workshop on ontologies and information sharing, pages53–62, 2001.
Michel Klein. Change Management for Distributed Ontologies. PhD thesis, VrijeUniversiteit Amsterdam, Amsterdam, The Netherlands, 2004.
Mariem Mahfoudh, Germain Forestier, Laurent Thiry, and Michel Hassenforder.Consistent ontologies evolution using graph grammars. In Knowledge Science,Engineering and Management, pages 64–75. Springer, 2013.
Salvatore Raunich and Erhard Rahm. Towards a benchmark for ontology merging. InOn the Move to Meaningful Internet Systems: OTM 2012 Workshops, pages124–133. Springer, 2012.
Grzegorz Rozenberg. Handbook of graph grammars and computing by graphtransformation, volume 1. World Scientific, 1999.
M.Mahfoudh et al. Merging Ontologies with SPO 24/25
Context Typed Graph Grammars Approach Implementation Conclusion References
Thank you for your attention
Contact: [email protected] site: www.mariem-mahfoudh.info
M.Mahfoudh et al. Merging Ontologies with SPO 25/25