Content uploaded by Fabio Santos
Author content
All content in this area was uploaded by Fabio Santos on Dec 05, 2018
Content may be subject to copyright.
A Proposal for Optimizing Internetworking
Matching of Network of Ontologies
Fábio Santos, Kate Revoredo, Fernanda Baião
Department of Applied Informatics
Federal University of the State of Rio de Janeiro (UNIRIO), Brazil
{fabiomarcos.santos, katerevoredo, fernanda.baiao}@uniriotec.br
Introduction
Future Work
The work was partially funded by UNIRIO (PQ-UNIRIO N01/2018 ) and CNPq (401505/2014-6)
•A network of ontologies is formally defined as Γ=< Ω, Λ>, where Ω
is a finite set of ontologies and Λ(O,O′)is aset of alignments
between pairs of ontologies belonging to Γ.
•Given a set of two or more networks of ontologies Ψ={Γ1,Γ2, ...,
Γn}, the internetwork matching problem searches for a final network
of ontologies Γfresulting from the alignments of the networks in Ψ.
•Possible approaches are pairwise and holistic but both need to
compute all possible entities from each network. They do not
handle:
•Isomorphisms: identical correspondences between same entities -
in Figure 1 A1,1′={<O1.a1,O′1.a′′1,=> ,< O1.b1,O′1.b′′1,=>}).
•Trivial alignments: group of entities, that was previously aligned in
a network, appears in another network. In Figure 1 - A1,2 =<
O1.b1,O2.d2,⊑> and A1′,2′=< O′1.b′1, O′2.d′2,⊑> will work unnecessarily
to produce A1,2′=< O1.b1, O′2.d′2,⊑> and A1′,2 =< O′1.b′1, O2.d2,⊑>
Problem Definition
Approach and Preliminary Results
•All pairs of entities from each ontology that composes the networks
are analyzed, which poses a severe restriction in terms of
scalability.
• A System-of-Systems (SoS) is defined as a set of independent
information systems (IS), providing functionalities derived from the
interoperability among them [3].
Considering• that each IS within a SoS is conceptually described by
a unique ontology describing its domain, the conceptual support for
the whole SoS demands the interoperation of its composing IS,
thus requiring the alignment of all the corresponding ontologies.
(Figure 1).
This• work proposes an optimized approach for the internetwork
matching challenge [1] that tries to reduce the number of pairs to
be evaluated during the matching process, thus avoiding unneeded
computation while preserving the alignment quality.
Adapting,
Learning and
Integrating
Conceptualizations
Environment
Experiment Pairwise InterSubNM % of Reduction
2x2 14,138 5,608 60.3
3x2 22.236 10,027 54.9
3x3 38.893 27,039 30.4
4x3 42,319 27,039 36.1
4'x357,497 43,420 24.4
http://www2.uniriotec.br/ontologyalignment
•We proposed SubInterNM that avoids unnecessary computation by
identifying and reusing trivial alignments already computed in the
networks of ontologies that are being aligned and isomorphisms.
•Applying algebraic operations defined in [5] we can avoid
isomorphisms and consequently reduce comparison costs:
• Ω as O1∪O2…∪On− O1∩ O'1− O1∩ O'2−...−On∩ O'n
• Ω' as O'1∪O'2…∪O'n− O'1∩ O1− O'1∩ O2−...−O'n∩ On
• 2x2: Ω = {conference, cmt} and Ω′ = {cmt, sigkdd};
3x2: Ω = {conference, cmt, ekaw} and Ω′ = {cmt, sigkdd};
3x3: Ω = {conference, cmt, ekaw} and Ω′ = {cmt, sigkdd,
conference};
4x3: Ω = {conference, cmt, dblp, ekaw} and Ω′ = {cmt, sigkdd,
conference};
4’x3: Ω = {conference, cmt, edas, ekaw} and Ω′ = {cmt, sigkdd,
conference}.
Table 1: Total number of comparisons computed by each approach
Figure 1: Two networks of ontologies with previous and internetwork alignments
•We are planning investigate how use the previous alignments to
reduce even more the comparison costs in a internetwork matching.
•We are also planning check if it is possible to reduce costs to
compute the closures when two or more networks are aligned,
instead of competing everything from the scratch.
[1] - Santos, F., Revoredo, K., Baiao, F., Network of Ontologies – A Systematic Mapping Study and Challenges Comparison,
Technical Report. Relate-DIA/UNIRIO, RT-0005/2017, 2017.
http://www.seer.unirio.br/index.php/monografiasppgi/article/view/6833
[2] - de Abreu Santos, F.M., Revoredo, K., Bai ̃ao, F.A.: Paving a research roadmap on network of ontologies. In:
Proceedings of the 12th International Workshop on Ontology Matching co-located with the 16th International Semantic Web
Conference (ISWC 2017), Vienna, Austria, October 21, 2017. pp. 221–222 (2017), http:// ceur- ws.org/Vol-
2032/om2017_poster8.pdf
[3] - Boehm, B.: A view of 20th and 21st century software engineering. In: Proceedings of the 28th international conference
on Software engineering. pp. 12–29. ACM (2006)
[4] - Euzenat, J.: Revision in networks of ontologies. Artificial intelligence 228, 195–216 (2015),
ftp://ftp.inrialpes.fr/pub/exmo/publications/euzenat2015a.pdf
[5] - Casanova, M.A., de Macedo, J.A., Sacramento, E.R., Pinheiro, Aˆ.M., Vidal, V.M.,
Breitman, K.K., Furtado, A.L.: Operations over lightweight ontologies. In: OTM Confederated International Conferences” On
the Move to Meaningful Internet Systems”. pp. 646–663. Springer (2012)
[6] - Casanova, M. A., & Magalhães, R. (2018). An Algebra of Lightweight Ontologies. arXiv preprint arXiv:1809.01621.
•Let O1=(V1,Σ1) and O2=(V2,Σ2)be two ontologies, W be a subset
of V1, and Ψbe aset of constraints in V1.[5]
•The deprecation of Ψfrom O1=(V1,Σ1), denoted σ[Ψ](O1), returns
the ontology OD=(VD,ΣD), where VD= V1and ΣD=Σ1-Ψ. [6]
•The projection of O1=(V1,Σ1) over W, denoted π[W](O1), returns
the ontology OP=(VP,ΣP), where VP= W and ΣPis the subset of the
constraints in τ[Σ1] that use only classes and properties in W. [5]
•The union of O1=(V1,Σ1) and O2=(V2,Σ2), denoted O1∪O2, returns
the ontology OU=(VU,ΣU),where VU=V1∪V2and ΣU=Σ1∪Σ2. [5]
•The intersection of O1=(V1,Σ1) and O2=(V2,Σ2), denoted O1∩O2,
returns the ontology ON=(VN,ΣN), where V2= V1∩V2 and ΣN=
τ[Σ1]∩ τ[Σ2]. [5]
Operations Over Ontologies
•The difference of O1=(V1,Σ1) and O2=(V2,Σ2), denoted O1- O2,
returns the ontology OF=(VF,ΣF), where VF= V1and ΣF=τ[Σ1] -
τ[Σ2]. [5](Figure 2)
Figure 2: lightweight operations example [6]
