This tutorial tries to answer the following questions:
What is the best practice for ontology reuse?
Is it fine to use external ontology entities to model my local entities?
Should I import the ontologies that I reuse?
What if I only need a part of an ontology?
What if an external ontology that I reused, changes?
1. 1
Methods for ODP reuse
Valentina Presutti
Semantic Technology Laboratory (STLab), ISTC-CNR
Rome/Catania, Italy
17/10/2016, Kobe, Japan
ODP for Linked Data Publishing @ ISWC 2016
http://www.slideshare.net/vpresutti/methods-for-ontology-design-patterns-reuse
2. • Questions we will try to answer
• Background on ontology reuse
• Classification of reuse models
• Impact of reuse models
• Reuse models with their advantages and disadvantages
• Conclusion towards working examples
Outline
3. What is the best practice for ontology reuse?
Is it fine to use external ontology entities to model my local
entities?
Should I import the ontologies that I reuse?
What if I only need a part of an ontology?
What if an external ontology that I reused, changes?
Questions we will try to answer
4. Ontology reuse is a recommended practice (see e.g. [Burleson et al.,
2014], [Bizer et al., 2009])
reuse of standard vocabularies
linked data principles
It favors semantic interoperability [Simperl, 2009]
knowledge reuse is a key success factor for the Semantic Web
reusability is an intrinsic property of ontologies
Ontology Design Patterns (ODP) [Gangemi, 2005]
they make reuse easier, as they isolate specific requirements
Ontology reuse
5. [d’Aquin and Noy, 2012]: how to choose ontology libraries and current
open issues for ontology library developers (many are still open)
[Suárez-Figueroa et al., 2011]: methodological guidelines for ontology
engineering. As for reuse it identifies two main scenarios and lists the main
activities that characterise them: reuse of ontologies as they are and reuse
by alignment. ODP reuse is also considered but the possible process is not
detailed in terms of activities
[Fernández-López et al, 2011]: guidelines on how to identify parts of
ontologies to be reused
[Schaible et al, 2014]: survey on vocabulary reuse strategies in linked data.
It shows that popularity of vocabulary is one of the most used criteria for
reuse and that the focus is at terminology reuse
This tutorial is mainly based on [Presutti et al, 2016], [Lodi et al, 2016],
and [Hammar and Presutti, 2016] which focus on ontology design patterns
reuse for linked open data publishing
Literature on ontology reuse
6. Non-linear evolution of ontology design
Diversity of modelling requirements
Availability of existing ontologies
Sustainability within organisations
Trends
Personal taste of ontology designers
No standard for reuse
8. • Type of reused ontologies
• Type of reused ontology fragments
• Amount of reused axioms
• Alignment policy
Classification of reuse models
9. Reuse of foundational [Gangemi et al, 2003] or top-level
ontologies [DBpedia Ontology]
Specialising DOLCE or the DBpedia Ontology
Modelling events and participation in them
Cultural institutes involved in an exhibition
Type of reused ontologies
10. Reuse of Ontology Design Patterns
Participation ODP to model event participation
Type of reused ontologies
11. Reuse of domain ontologies
Reusing the Event Ontology [Raimond and Abdallah, 2007]
Music events
Type of reused ontologies
12. Reuse of individual entities (classes, relations, individuals)
dolce:hasParticipant, dbpedia-owl:Event
Reuse of “groups” of entities (modules, ODP, or arbitrary
fragments)
Participation ODP
dolce:hasParticipant and all its relevantly related entities,
e.g. dolce:Object, dolce:Event, etc.
The whole Event Ontology
Type of reused ontology fragments
13. Reuse of ontologies including all their axioms
the whole DOLCE, the whole Event Ontology
Reuse only of axioms in a given neighborhood of a specific entity
referred to as ontology module
dolce:hasParticipant + entities and axioms within a certain
graph distance
Only reuse individual entities’ URIs with no axioms
Amount of reused axioms
14. Direct reuse
entities and axioms delegated to an external ontology
dolce:Event as type of ontology individuals in my ontology
Indirect or template-based reuse
define my own entities and align them to external ontologies
myont:CulturalEvent rdfs:subClassOf dolce:Event
XDP (WebProtégé plugin) offers tool support (afternoon
session)
Alignment policy
15. Reusing an ontology, ontology fragment, or ontology entity does
not imply to use <owl:import>, nevertheless
In order to assess semantic coherence and consistency of the
resulting ontology wrt its requirements, one needs to
<owl:import> the reused ontologies, at least for the time needed
to perform these tasks
<owl:import>
16. The only shared characteristic among all these practices is that
entities are reused with their original logical type
If Ol reuses Or then
e rdfs:type owl:Class and e ∈ Or
implies
f rdfs:type owl:Class and e ∈ Ol
e ≡ | ⊆ f
Reuse models
17. All these models can be mixed in a same ontology project, or
different projects from the same organisation may apply different
strategies
There is not “the best for all situations” model
Regardless the trend or taste, the type of reuse impacts on the
developed ontology project
The choice of reuse model must be done according to the
ontology project’s contextual requirements
Reuse models
18. Ontology semantics
maximal commitment: when we reuse a whole ontology with
all its axioms
minimal commitment: when we indirectly reuse an individual
entity without importing its related axioms
semantics is safer and more complete in case of maximal reuse
Impact of reuse models
19. Sustainability and usability
maximal reuse may mean less usability
useless or undesired entities and axioms
strong dependency on external resources
risk of incoherence wrt original requirements after external
changes
Impact of reuse models
20. Interoperability
minimal reuse may simplify interoperability
the less constraints given by axioms the simpler
interoperability
Impact of reuse models
21. The focus on quality of semantics pushes towards maximal reuse,
while the focus on interoperability and sustainability/usability pushes
towards minimal reuse
The vision of the designer, the scope of the project, the type of
project and the nature of data that we deal with are at the basis
of the choice of a reuse model
The goal is to maximise the quality of semantics without
negatively impact on usability, sustainability, and interoperability
Impact of reuse models
22. Direct reuse of individual entities
E
X
T
O
N
T
L
O
C
O
N
T
Advantages Disadvantages
Semantic ambiguity
Difficulty in verifying consistency
Strong external dependency
Risk of instability
Possible sustainability issues
Linked data praxis
Reuse of shared terminology
Good if one wants to comply
with and follow evolution of
standards
23. Direct reuse of ODP
2
3
Advantages Disadvantages
Dependency on external module
Mitigated risk of instability
Stability and sustainability
Modularity
Interoperability
ODP are unlikely to change
Easy to re-design in case of
changes
loc:Person
loc:organises
24. Indirect reuse of modules
2
Advantages Disadvantages
Possible heterogeneity in module
identification
Difficulty in providing formal
specification of external module
Effort for replicating the module
implementation
Dependency on external changes is
limited to alignment axioms
Easier re-design for fixing issues due
to external changes
Stability and sustainability
Modularity and Interoperability
Alignment
axioms
25. Indirect reuse of ODP
2
Advantages Disadvantages
Effort for replicating the ODP
implementation
Dependency on external changes is
limited to alignment axioms
Easier re-design for fixing issues due
to external changes
Stability and sustainability
Modularity and interoperability
Alignment
axioms
26. Summarising table
Direct reuse Indirect reuse
Individual
entities
• Dependency on standards,
if required
• LD praxis
• Shared vocabulary
• Less effort in design
• Semantics ambiguity
• Difficult to verify
consistency/coherence
• Strong external
dependency
• Risk of instability
• Possible sustainability
issues
• Dependency on standards
• LD principles
• Shared vocabulary
• Easier to fix possible issues
due to external changes
• Semantics ambiguity
• Indirect dependency from
external resources
• More sustainable
Module
• Less effort in design
• Clearer semantics
• Strong dependency on
external resources
• Possible heterogeneity in
module identification
• Difficulty in providing
formal specification of
external module
• Hard to fix possible
issues due to external
changes
• Limited dependency on
external changes (only
alignment axioms)
• Easier re-design for fixing
issues due to external
changes
• Stability and sustainability
• Modularity and
Interoperability
• Possible heterogeneity in
module identification
• Difficulty in providing
formal specification of
external module
• Effort for replicating the
module implementation
ODP
• Reuse of design good
practices
• Stability and sustainability
• Modularity and
interoperability
• ODP are unlikely to
change
• Easy to re-design in case
of changes
• Dependency on external
module
• Mitigated risk of
instability
• Dependency on external
changes is limited to
alignment axioms
• Easier re-design for fixing
issues due to external
changes
• Stability and sustainability
• Modularity and
interoperability
• Effort for replicating the
ODP implementation
27. The second part of the morning session (Pascal Hitzler and
Giorgia Lodi) is about use case scenarios
both academic and real world scenarios will be discussed and
different reuse models applied
The afternoon (Karl Hammar) will be dedicated to a practical
session where you will apply ontology reuse with some tools
support (XDP, WebProtégé plugin)
Working examples
28. • Ontology reuse is a key factor for the success of Semantic
Web technologies
• There is no “one fits all” solution
• Tradeoff between semantic ambiguity, usability, sustainability
and interoperability, is the goal
• The stronger the commitment the safer and more complete
the semantics, the lower the usability/sustainability and
interoperability
• Different reuse models with their advantages and
disadvantages
Conclusion
30. [Bizer et al., 2009] Christian Bizer, Tom Heath, Tim Berners-Lee:
Linked Data - The Story So Far. Int. J. Semantic Web Inf. Syst. 5(3): 1-22 (2009)
[Burleson et al., 2014] Linked Data Platform Best Practices and Guidelines.
W3C Working Group Note 28 August 2014 https://www.w3.org/TR/ld-bp/
[Simperl, 2009] Elena Paslaru Bontas Simperl: Reusing ontologies on the Semantic Web:
A feasibility study. Data Knowl. Eng. 68(10): 905-925 (2009)
[Gangemi, 2005] Aldo Gangemi: Ontology Design Patterns for Semantic Web
Content. International Semantic Web Conference 2005: 262-276
[Gangemi et al. 2003] Aldo Gangemi, Nicola Guarino, Claudio Masolo, Alessandro
Oltramari: Sweetening WORDNET with DOLCE. AI Magazine 24(3): 13-24 (2003)
[d’Aquin and Noy, 2012]
[DBpedia Ontology] The DBpedia Ontology: http://wiki.dbpedia.org/services-
resources/ontology
[Raimond and Abdallah, 2007] The event ontology. Technical report, 2007.
References
31. [Suárez-Figueroa et al., 2011] Mari Carmen Suárez-Figueroa, Asunción Gómez-
Pérez, Mariano Fernández-López: The NeOn Methodology for Ontology
Engineering. Ontology Engineering in a Networked World 2012: 9-34
[Fernández-López et al, 2011] Mariano Fernández-López, Mari Carmen Suárez-
Figueroa, Asunción Gómez-Pérez: Ontology Development by Reuse. Ontology
Engineering in a Networked World 2012: 147-170
[Schaible et al, 2014] Johann Schaible, Thomas Gottron, Ansgar Scherp: Survey on
Common Strategies of Vocabulary Reuse in Linked Open Data
Modeling. ESWC 2014: 457-472
[Presutti et al, 2016] Valentina Presutti, Giorgia Lodi, Andrea Giovanni Nuzzolese, Aldo
Gangemi, Silvio Peroni and Luigi Asprino: "The role of Ontology Design Patterns in
Linked Data projects”. ER 2016.
[Lodi et al, 2016] Semantics for Cultural Heritage Valorisation. Giorgia Lodi, Valentina
Presutti, Luigi Asprino, Andrea Nuzzolese, Diego Reforgiato, Aldo Gangemi, Annarita
Orsini, Chiara Veninata. Springer, Data Analytics in Digital Humanities, 2016
[Hammar and Presutti, 2016] Karl Hammar and Valentina Presutti - Template-Based
Content ODP Instantiation. In WOP 2016. IOS Press
References
Editor's Notes
This approach consists on directly introducing individual entities of external ontologies in local axioms.
This practice is very common in the LD community, however it is a routine, not a good practice, at all.
It is essentially driven by the intuition of the semantics of concepts based on their names, instead of their axioms.
In this case, the risk that the formal semantics of the reused entities is incompatible with the intended
semantics to be represented is rather high. Moreover, with this practice a strong dependency of the local ontology
with all the reused ontologies is created. This dependency may put at risk the sustainability and stability of the local ontology
and its associated knowledge bases: if a change in the external ontology introduces incoherences in the local one,
they must be dealt with a redesign process and consequential change in the ontology signature.
If the fragment is clearly and formally identified, since it is embedded in a dedicated ontology,
some of the previous remarked issues can be mitigated. Let us consider
that the earlier example class ex:Event is defined in an external ontology that
implements a specific ODP. In this case, a scenario in which a redesign process
must be undertaken may be less frequent. In fact, ODPs are developed for reuse
purposes and thus they are unlikely to change. In the light of these observations,
it is recommended to reuse ODPs in contrast to individual entities.
With this approach, the modelling of some concepts and relations, which are relevant for the
domain but applicable to more general scopes, is delegated to external ontologies by means of ontology module reuse. An ontology module is a fragment that may be identified as providing a solution to one or more specific requirements of the local ontology. For example, let us consider an external ontology modelling the participation of an individual (e.g. through a property ex:isInvolvedIn ) to an
event (e.g. a class ex:Event ). If the local ontology needs to specify a particular involvement in an event (e.g. lo:hosted ) it should specialise (it indirectly reuses) the relation of the external one (i.e. ex:isInvolvedIn ). The fragment of the external ontology identified as relevant for the local ontology may be communicated in some usage documentation provided with the ontology. Nevertheless, it is difficult to provide third parties with a formal indication of the fragment that was meant to be relevant. This may lead to high heterogeneity in the usage of external fragments in data modelled through the local ontology. As for ontology sustainability, when a change in the external ontology provokes possible incoherences, the redesign process would be easier dealt with as compared to the previous approach.