=Paper=
{{Paper
|id=Vol-1517/JOWO-15_WoMO_paper_2
|storemode=property
|title=A Novel Approach for Extracting Well-Founded Ontology Views
|pdfUrl=https://ceur-ws.org/Vol-1517/JOWO-15_WoMO_paper_2.pdf
|volume=Vol-1517
|dblpUrl=https://dblp.org/rec/conf/ijcai/LozanoCA15
}}
==A Novel Approach for Extracting Well-Founded Ontology Views==
https://ceur-ws.org/Vol-1517/JOWO-15_WoMO_paper_2.pdf
A novel approach for extracting well-founded ontology views
Jose Lozano and Joel Carbonera and Mara Abel
Institute of Informatics
Universidade Federal do Rio Grande do Sul
Porto Alegre Brazil
Abstract how some aspects of the ontology are considered during
the extraction process. We also carried out an experiment
When the size of an ontology increases, it becomes hard
for demonstrating that our approach produces WFOVs that
to be managed. Ontology view extraction is an approach
that can be used for overcoming the challenges that arise are smaller and that fit better to their target conceptualiza-
in this scenario. In this context, an ontology view is a tions than the WFOVs extracted by the original approach
subset of an ontology tailored to a specific set of user re- (Lozano et al. 2014). This experiment was based on a data-
quirements. Well-founded ontology views were recently driven method for evaluating approaches for ontology view
proposed as ontology views that follow well-founded extraction. This method is based on comparisons of the f-
ontological principles, which ensures some desirable measures of different ontology views, considering sets of
ontological properties. In this paper, we propose a novel terms extracted from the scientific literature related to dif-
approach for extracting well-founded ontology views, ferent communities or tasks.
which is more flexible than the previous approach. We In Section , we provide an overview of the main ap-
also present a method for evaluating the quality of ap-
proaches available in the literature for extracting portions
proaches for extracting ontology views. We apply this
method for demonstrating that our novel approach pro- of ontologies. In Section , we present a basic definition
duces ontology views that are more accurate than those of the notion of well-founded ontology view and describe
produced by the previous approach. We illustrate our the basic approach for extracting WFOVs. Section presents
approaches using a domain ontology for Petrography. our approach for extracting WFOVs. Section describes the
method that we used for evaluating our approach. Section
describes the application of the different approaches for ex-
Introduction tracting WFOVs in a case scenario with their corresponding
Ontologies tend to evolve over time by incorporating new evaluations. Finally, Section presents our conclusions.
knowledge. The resulting ontology can lead to a scenario
of information overload, where the information exceeds the Related Works
cognitive capability of the users. Ontology views have been
In general, the literature provides two main approaches that
adopted as a solution for overcoming this scenario, since
can be used for extracting manageable portions of ontolo-
they are extracted from a base ontology according to spe-
gies. The extraction of ontology modules (Doran, Tamma,
cific user criteria, and provide only the knowledge that is
and Iannone 2007; d’Aquin, Sabou, and Motta 2006; Sei-
relevant for a given task at hand.
denberg and Rector 2006) fragments a given base ontol-
The literature provides some approaches for extracting
ogy into a set of smaller, non-overlapping and possibly in-
ontology views (Noy and Musen 2003; Bhatt et al. 2004;
terconnected parts, or modules. The alternative approach,
Lozano et al. 2014). Particularly, in (Lozano et al. 2014), the
is the extraction of ontology views (Noy and Musen 2003;
authors propose the notion of well-founded ontology view
Bhatt et al. 2004), where smaller (and possibly overlapping)
(WFOV), which is an ontology view that preserves some
subsets of the base ontology are extracted according to the
important ontological meta-properties (such as identity and
user requirements. Since they are tailored to specific tasks
existential dependence). The authors also define a set of
or interests, ontology views provide to the agent (users or
conservation principles and apply them for guiding a sub-
computer applications) only the knowledge that is relevant
ontology extraction algorithm.
for reaching some goal.
In this paper, we propose a new approach for extract-
Some of these approaches (Seidenberg and Rector 2006;
ing WFOVs, which modifies the basic approach defined in
d’Aquin, Sabou, and Motta 2006; Noy and Musen 2003)
(Lozano et al. 2014). Our novel approach eliminates a source
are dependent on some representation language (such as
of information overload from the basic approach and pro-
OWL), while others (Doran, Tamma, and Iannone 2007;
vides more flexibility, since it allows the user to specify
Bhatt et al. 2004), language-independent, adopt an abstract
ontology representation that is based on graphs. Besides
that, most of the approaches extract modules or views start-
ing from some target concepts and include in the subset Well-founded Ontology Views
(module or view) only the ontology elements (concepts, re- In this Section, we present the approach proposed by
lations and properties) that are directly related to the con- (Lozano et al. 2014), for extracting well-founded ontology
cepts that are already included in the subset. views (WFOV). Since this approach relies on a set of on-
tological meta-properties, firstly we shall discuss them. Af-
Algorithm 1 The basic approach for WFOV extraction. ter, we present the characterization of a WFOV. Finally, we
Require: Well-Founded Ontology present the basic approach for extracting WFOVs, proposed
procedure SEL(Ob , tConcepts, tRelations, So ) by the authors.
So .C So .C [ tConcepts
So .R So .R [ tRelations Ontological Meta-Properties
newC ;
The approach proposed by (Lozano et al. 2014) uses the for-
newR ;
for all c 2 tConcepts do
mal characterization of the ontological meta-properties pro-
conservesT AX(Ob , c, newC, newR)
vided by the Unified Foundational Ontology (UFO) (Guiz-
conservesQU A(Ob , c, newC, newR) zardi 2005). This ontology provides a set of categories of
conservesIP (Ob , c, newC, newR) universals, which are characterized according to a set of
conservesED(Ob , c, newC, newR) meta-properties. The categories of universals can be viewed
conservesRD(Ob , c, newC, newR) as meta-types, since they are types of types. Thus, they can
conservesF R(Ob , c, newC, newR) be used for classifying classes in specific domain ontologies.
conservesP R(Ob , c, newC, newR) When some class C is classified by some meta-type M T ,
newC newC So .C this means that C has the meta-properties that characterize
newR newR So .R
M T , and this entails some formal consequences, accord-
end for
if newC 6= ; then
ing to the UFO axiomatization. The UFO has been used for
SEL(Ob , newC, newR, So )
supporting the development of domain ontologies (Carbon-
else era et al. 2011; 2013; Carbonera, Abel, and Scherer 2015;
if newR 6= ; then Abel, Perrin, and Carbonera 2015) in a well-founded basis.
So .R So .R [ newR Here we will present the main meta-properties and meta-
end if types provided by UFO and that are used by the approach
end if of (Lozano et al. 2014). A detailed account of UFO can be
end procedure found in (Guizzardi 2005).
One of the main categories of universals provided by UFO
In (Lozano et al. 2014), the authors propose using onto- is Substantial Universal, whose instances are individuals
logical meta-properties (such as identity, rigidity and exis- that, in general, are existentially independent of all other
tential dependency) for guiding the extraction of ontology individuals. Some of its instances can be existentially de-
views. Their approach has the advantage of including in the pendent when they are considered inseparable parts of their
views the ontology elements (concepts, relations and prop- hosts. Sortal Universals are substantial universals that pro-
erties) that need to be included in the view due to their onto- vide or carry some principle of identity (PI) for their in-
logical status. For example, if the concept A is included in stances. In this context, a PI is the principle that supports
an ontology view and instances of A are existentially depen- the judgment whether two instances of the universal are the
dent on instances of a concept B, B should also be included same.
in the view. This dimension of analysis is not considered Another important ontological meta-property used by
by the other approaches discussed in this section. Since our UFO is the rigidity. A certain universal is rigid when its
work proposes an improvement of the approach proposed extension (set of all particulars) is the same in all possible
by (Lozano et al. 2014), in the Section we shall present this worlds. That is, an instance of a rigid universal cannot cease
approach in more details. to be an instance of it without ceasing to exist. For example,
In Table 1, we present a comparison of the approaches Person can be viewed as a rigid universal, since persons can-
discussed in this section. The approaches are identified as: 1 not cease to be persons without ceasing to exist; meanwhile
(d’Aquin, Sabou, and Motta 2006), 2 (Doran, Tamma, and all instances of Student (which is an anti-rigid universal) can
Iannone 2007), 3 (Noy and Musen 2009), 4 (Seidenberg and still exist (as persons) if they cease to be students.
Rector 2006), 5 (Bhatt et al. 2004) and 6 (Lozano et al. Within the sortal universals, UFO includes three distinct
2014). types of substance sortals, which are rigid sortals that pro-
vide their own principle of identity: Kind, which repre-
Table 1: Comparison of sub-ontology extraction approaches. sents functional complexes (Person, Dog, Chair, etc); Col-
lective (Swarm, Forest, etc), which represents collectives;
Approach 1 2 3 4 5 6
and Quantity, which represents objectified portions of mat-
Language- No Yes No No Yes Yes
independent
ter (Wine, Water, Gold, etc). Besides that, Subkind is a rigid
Use of Meta- No No No No No Yes sortal that does not provide its own PI, but carries a princi-
properties ple of identity that is supplied by a given substance sortal.
Context Module Module View Module View View UFO also defines two anti-rigid sortals: Roles and
Phases. Phases are universals that constitute possible stages
in the history of a substance sortal. Phases are relationally Figure 1: A WFOV for the Diagenesis community, extracted
independent, since they depend solely on intrinsic proper- from a domain ontology for Petrography
ties. For example, Baby, Toddler, Kid, Teenager and Adult
are considered phases of Human. On the other hand, Roles
are relationally dependent, since they depend on extrinsic
(relational) properties. This is the case, for example, when
we say that for an instance of person to be considered a Stu-
dent, she must be enrolled at an educational institution.
Other substantial universals do not have the properties of
sortals; they are dispersive universals. This is the case, for
example, of Categories, which are rigid universals that do
not provide or carry a PI for their instances. Categories rep-
resent essential properties that are common to all instances
of many disjoint universals that provide distinct PIs. Ratio- ponentOf, memberOf, subCollectionOf and subQuantityOf.
nal agent is an example of Category, since it abstracts an es- Each parthood relation can only be established between in-
sential property (namely, the rationality) of instances of Per- dividuals of specific UFO meta-types, respecting some on-
son and Artificial Agent, which are disjoint universals, with tological constraints embedded in UFO. These relations can
distinct PIs. Role Mixins, on the other hand, are anti-rigid be characterized by five meronymic meta-properties that in-
universals that do not provide and do not carry a PI for their dicate: essential part, inseparable part, immutable part, im-
instances. They can be viewed as generalizations of roles of mutable whole and shareable part.
different substance sortals. For example, Customer is a role As important as the characterization of the meta-
mixin that generalizes Personal Customer, which is a role of properties and meta-types are, UFO also provides some pos-
Person; and Corporate Customer, which is a role of Organi- tulates that a model should follow:
zation. Finally, Mixins are universals that do not provide and • Postulate 1: Every individual in a conceptual model of
do not carry a PI for their instances and that are semi-rigid; the domain must be an instance of a sortal.
that is, they have some instances that are necessarily their
instances, but they also have some instances that are only • Postulate 2: An individual represented in a conceptual
contingently their instances. They usually generalize rigid model of the domain must instantiate exactly one ultimate
and anti-rigid universals. For example, Seatable Object is a Substance Sortal (kind, quantity or collective).
mixin that generalizes Chair, which is a rigid universal; and • Postulate 3: A rigid universal cannot specialize (restrict)
Solid Crate, which is an anti-rigid universal (actually, it is a an anti-rigid one.
phase of a Crate, which can also be a Broken Crate).
On the other hand, Moment Universals are Universals • Postulate 4: A dispersive universal cannot specialize a
whose instances are existentially dependent individuals that Sortal.
inhere in other individuals. Some moment universals depend Furthermore, it is important to notice that every sortal that
existentially on a single entity. This is the case of Qual- does not provide its own principle of identity (Role, Phase
ity Universals and Modes. Quality Universals represents and SubKind) must be subsumed by exactly one concept that
the properties in the conceptual models. A Quality Univer- provides its own identity (one of the Substance Sortals).
sal characterizes other Universals and is related to Quality
Structures, that is, a structure that represents a set of all val- Basic Definitions
ues that a quality can assume. Thus, considering the prop-
The notion of ontologically well-founded ontology view
erty Color as a Quality Universal, a given instance of Car
is defined considering certain principles of conservation
could be characterized by an instance of Color, which is as-
proposed by (Lozano et al. 2014). These principles were
sociated with a value (called quale) in the ColorStructure,
built considering a set of philosophically well-founded on-
which represents all the possible values that the property
tological meta-properties. In this work, the selected meta-
Color can assume. On the other hand, Modes are universals
properties were obtained from the UFO ontology.
whose instances are existentially dependent individuals, and
that are not associated to Quality Structures. Examples of In order to illustrate the proposed conservation principles,
modes are Skill, Belief, Headache, etc. Both Quality univer- we present portions of two WFOVs generated from a base
sals and Modes are related to the entities that they character- ontology for the domain of Petrography (a field of Geol-
ize through a relation of characterization. Besides that, Re- ogy). These WFOVs were generated for meeting the inter-
lators are moments that depend existentially on two or more ests of two different communities of users within the domain
entities. Examples of relators are Enrollment, Contract, etc. of Petrography: Diagenesis (Figure 1) and Microstructural
Relators are related to entities that it relates through a rela- Analysis (Figure 2).
tion of mediation. The relators also represent the relational The principles of conservation proposed by (Lozano et al.
dependency of roles and role mixins. Due to this, roles and 2014) are:
role mixins must be related to some relator, through a rela- Conservation of identity: If a view v includes a concept c
tion of mediation. that does not provide its own principle of identity, then
UFO proposes four types of parthood relations: com- v should also include all the supertypes of c from which
Figure 2: A WFOV for the community of Microstructural Figure 3: New version example
analysis, extracted from a domain ontology for Petrography
(a) If it were applied the approach 1 in the target concept,
the concepts colored in light gray are included in the view.
(b) If were applied the new version (approach 2)
c inhered its principle of identity, as well as, all the sub-
sumption relations that are held between these concepts.
For example, if zeolite is included in v, mineral should
also be included, since mineral provides the identity to
zeolite.
Conservation of the existential dependence: If a concept
c1 is included in the view v, and instances of c1 are exis-
tentially dependent on instances of c2 , then it is necessary
to include in v also the concept c2 and the relation held
between c1 and c2 . For example, if Porosity is included in
v, the concept Rock must be included because the porosity
is existentially dependent of Rock.
Conservation of relational dependence: If a concept c1 is
included in the view v, and c1 is relationally dependent of c1 should be included. For example, if the concept Rock
on a relation (materialized through a given relator) with Unit is included in v, the concepts deformation zone and
the concepts in {c2 , ..., cn }, then it is necessary to include sedimentary facies should also be included.
in v also: the relator r, all the concepts in {c2 , ..., cn }
and all relations that are held between the concepts in Basic Approach for extracting WFOV
{c2 , ..., cn }, r and c1 that are necessary for the conser- The basic approach for extracting WFOVs is formalized in
vation of the relational dependence. For example, if the the Algorithm 1. It takes as input the following parame-
concept is Cement is included in v, the concepts pore and ters: the ontology base (Ob ), a set of user required concepts
filling should also be included in v, because a mineral is (targets), a set of relations (relations), and the resulting
considered cement when the mineral is filling pore. Thus, extracted sub-ontology (So ). At the beginning, relations
cement is relational dependent of filling and pore. and So are empty. The algorithm analyses each concept in
Conservation of taxonomy: If a view v includes the con- targets. For each concept, the conservation principles are
cept c1 , it should also include all the concepts that are applied for ensuring that the result will be an ontologically
subsumed by c1 . For example, if concept Silicate Mineral well-founded ontology view. The conservation principles
is included in v, all the concepts that it subsumes are in- are applied through the following functions: conservesTAX,
cluded in the v. for the conservation of taxonomy; conservesQUA, for the
conservation of attributes; conservesIP, for conservation of
Conservation of attributes: If a view v includes a con- identity principle; conservesED, for the conservation of ex-
cept c1 , every attribute1 of c1 must also be included in istential dependence; conservesRD, for the conservation of
v. For example, if the concept Diagenetic Constituent is relational dependence, conservesFR, for the conservation of
included in v, the concept habit should also be included, formally related concepts; and conservesPR, for the conser-
because it is a quality of Diagenetic Constituent. vation of partonomy. In the main loop, these functions accu-
Conservation of formally related concepts: If a view in- mulate concepts (in newC) and relations (in newR) that are
cludes a concept c1 , every concept that is related to c1 in a necessary for ensuring the defined principles for a given con-
formal relation is added. For example, if the concept Rock cept c in tConcepts. More details regarding this approach
Unit is included in v, the concept Rock is also included be- can be found in (Lozano et al. 2014).
cause there is the formal relation constituted by between
Rock Unit and Rock. A novel approach for extracting well-founded
Conservation of partonomy: If a view includes a concept ontology views
c1 , all the concepts whose instances are parts of instances
Our new approach aims at reducing the number of concepts
1
Adopting the UFO, attributes are considered Quality Univer- included in the views by the basic approach, and covering
sals more precisely the requirements of a task at hand. In an
overview, our approach relaxes some criteria adopted by the the basic approach would include in the resulting view all
basic approach and provides more flexibility to the user. the concepts in gray. On the other hand, our novel approach
The main difference in our approach regarding the basic applies the conservation of taxonomy only to the target con-
approach concerns the application of the principle of con- cept (Grain), including the concepts surrounded by a circle
servation of taxonomy. While the basic approach includes in in Figure 3 (b). As a consequence, the sub-ontology will not
the resulting view the taxonomy of every concept that is al- include the taxonomy of Mineral and Intracrystalline Defor-
ready included in the view, our novel approach includes only mational Structure, depicted in Figure 3 (b) (in white).
the immediate taxonomy of the original target concepts. This
modification was motivated by the fact that, in general, only Algorithm 3 Parameterizable algorithm for selecting ontol-
the taxonomies of the target concepts are useful for the task ogy elements for the view
for which the view was built. Besides that, the inclusion of Require: Well-Founded Ontology
the taxonomies of every concept in the view leads to a rapid procedure SELECTION(Ob , tConcepts, tRelations, So , wP, wF R)
increase in the size of the view. In this way, the inclusion So .C So .C [ tConcepts
of irrelevant taxonomies can be considered as a source of So .R So .R [ tRelations
information overload. newC ;
Moreover, our approach also provides more flexibility to newR ;
for all c 2 tConcepts do
the user, by allowing the setting of three parameters. These
conservesQU A(Ob , c, newC, newR)
parameters are the variables wP (with Partonomy), wRT
conservesIP (Ob , c, newC, newR)
(only Rigid Taxonomy) and wFR (with formal relation). In conservesED(Ob , c, newC, newR)
this way, the ontology engineer can specify if the desired conservesRD(Ob , c, newC, newR)
WFOV should include the partonomies of every concept if wF R then
or not; if it should include only the rigid concepts in the conservesF R(Ob , c, newC, newR)
taxonomies or if non-rigid concepts should be included as end if
well; and if it should include all the concepts that are related if wP then
(through formal relations) to concepts already included in conservesP R(Ob , c, newC, newR)
the view. end if
newC newC So .C
Algorithm 2 Novel algorithm for extracting well-founded newR newR So .R
end for
ontology views
if newC 6= ; then
Require: Well-Founded Ontology selection(Ob , newC, newR, So , wP, wF R)
procedure E XTRACTOR(Ob , tConcepts, tRelations, So , wP, wRT, wF R) else
So .C So .C [ tConcepts if newR 6= ; then
newC ; So .R So .R [ newR
newR ; end if
for all c 2 tConcepts do end if
if wRT then end procedure
conservesT AXR(Ob , c, newC, newR)
else
conservesT AX(Ob , c, newC, newR)
end if Evaluation Method
end for We assume that the quality of the ontology view extraction
newC newC [ tConcepts approach can be measured by the degree to which the ex-
newR newR [ tRelations tracted ontology views fit to the required conceptualizations.
selection(Ob , newC, newR, So , wP, wF R)
In our work, we adopt an approach for evaluating this fitness
end procedure
in an indirect way.
We assume that the required conceptualization (of a com-
Our approach is formalized in the algorithm 2. Firstly, it munity, of some task) is properly represented, in natural
applies the conservation of taxonomy only to the original language, in the relevant literature. Thus, our evaluation is
target concepts. At this point, the parameter wRT controls if based on measuring the correspondences between a given
the taxonomy takes all the concepts or only the rigid ones ontology view (built for some specific task or some commu-
(Algorithm 4 presents how to recover only rigid concepts in nity) and the set of terms extracted from the relevant litera-
the taxonomy). Then, the algorithm calls the selection al- ture (related the correspondent task or community for which
gorithm (algorithm 3), which applies the other principles of the ontology view was built). We measure these correspon-
conservation, according to the parameters. Notice that the dences through well-known measures used in Information
algorithm 3 is a variation of the basic algorithm proposed by Retrieval: Precision, Recall and F-measure (Powers 2011).
(Lozano et al. 2014) (presented in subsection ), which does Considering this, we assume that in a useful approach for
not apply the conservation of the taxonomy of every concept extracting ontology views, an ontology view generated for
in the main loop, and which controls through parameters the a community A should have a value of f-measure that is
application of some principles of conservation. For instance, greater than the f-measure value of any well-founded on-
in Figure 3 (a), if we consider Grain as the target concept, tology view generated for community B, when compared to
Figure 4: Evaluation Method applied in the Petrography do- performed, following the sequence of steps defined in (Abel
main. 2001): (i) exclude all common words: prepositions, articles,
adverbs and connection verbs and; (ii) mark all geological
terms specific to the domain.
Algorithm 4 Conserve Taxonomy only Rigid
procedure CONSERVES TAXR(Ob , c, newC, newR)
for all v 2 Ob .C|9r = Rel(subsumption, c, v) do
if metaT ype(v) 2 {SubKind, Collective, Kind, Quantity, Category}
then
newR newR [ r
newC newC [ v
conservesT AXR(Ob , v, newC, newR)
end if
end for
a set of terms extracted from the literature of the commu- end procedure
nity A. In other words, a view generated for the community
X should fit better the conceptualization of the community X This extraction was done manually, for ensuring the qual-
rather than the ontology view generated for another commu- ity of the extraction. We excluded the terms that were not
nity. exclusive to the communities of Diagenesis and Microstruc-
Considering LT as the set of terms extracted from the lit- tural analysis. We also excluded the terms that were com-
erature, the precision (P) of the generated ontology view O mon for both communities. The result was two lists of ge-
is given by ological terms; one (DT ) for the community of Diagenesis
|OT \ LT |
P (OT, LT ) = (1) and other (M T ) for the community of Microstructural anal-
|OT |
ysis.
, the recall (R) is given by The next step consists of generating the two well-founded
|OT \ LT | ontology views (one for each community), from a set of key
R(OT, LT ) = (2) terms that are representative of the community. These key
|LT |
terms were provided by domain experts. The WFOVs that
, and the f-measure (F) is given by were considered in this evaluation were extracted using De-
P (OT, LT ) ⇤ R(OT, LT ) trital Constituent, Diagenetic Constituent and Pore for the
F (OT, LT ) = 2 ⇤ (3)
P (OT, LT ) + R(OT, LT ) WFOV of Diagenesis and Deformational Band, Fault, Brec-
cia and Microfracture for the community of Microstructural
where |S| indicates the cardinality of the set S and OT is analysis.
the set of terms that identify the set of ontology elements The result of this step is an ontology view for Diagen-
(concepts, relations and properties) of the ontology view O. esis (DO) and ontology view for Microstructural analysis
(M O). For our proposed approach, we extracted one WFOV
Evaluation Results for each community, considering each combination of pa-
This section describes the application of the evaluation rameters.
method described in Section for comparing the performance The last step is to calculate and compare the f-measures,
of our approach (A2) for extracting ontology views with the considering the WFOVs and the sets of selected terms.
performance of the basic approach proposed by (Lozano et As depicted in Figure 4, we expect that the f-measure
al. 2014). In our evaluation, we compared the f-measure of (F DD = F M easure(DO, DT )) between ontology view
the ontology view generated by the basic approach A1 with for Diagenesis (DO) and the terms for the community of
the ontology views generated for each parameter combina- Diageneis (DT) is greater than the f-measure (F DM =
tion of our approach (A2). For performing this comparison, F M easure(M O, DT )) between ontology view for Mi-
we considered two WFOVs extracted from the domain on- crostructural analysis MO and the terms for the community
tology of Petrography proposed by (Lozano 2014): a WFOV of Diageneis DT. And, in the same way, it is also expected
for the community of Diagenesis and a WFOV for the com- that the f-measure (F M M = F M easure(M O, M T )) be-
munity of Microstructural analysis. These two communities tween MO and MT is greater than the f-measure (F M D =
of users employ different sets of concepts from the ontology F M easure(DO, M T )) between DO and MT. In the fol-
of Petrography. This base ontology of Petrography includes lowing subsections, we describe the evaluation of these two
366 concepts and 387 relations. cases. In Case 1, we evaluate the ontology view for Diagene-
Our approach also requires the extraction of sets of terms sis, comparing it with the ontology view for Microstructural
that are representative for the considered communities. For analysis, considering the terms for Diagenesis. In Case 2,
this step, a domain expert selected six peer-reviewed papers we evaluate the ontology view for Microstructural analysis,
about Diagenesis, such as (Worden and Burley 2003), and comparing it with the ontology view for Diagenesis, based
six papers about Microstructural analysis, such as (Haer- on the microstructural terms.
tel and Herwegh 2014). After the extraction of terms was In Table 2, we present the results of the evaluation process
of approaches A1 (basic approach) and A2 (novel approach), The row Case2 of Table 2 presents the ratio between the
in the two considered cases. Notice that the table presents measures of the ontology view of Microstructural analysis
the ratios between the considered measures of the two on- and the measures of the ontology view of Diagenesis. In this
tology views, evaluated according to a set of terms. Thus, row, it can be seen that approach A2 achieved its best result
in the row Case 1, the table presents, for both approaches using the parameter wP.
A1 and A2, the ratios between the measures (precision, re- The approach A2 achieved results of low quality in some
call and f-measure) of the ontology view for Diagenesis and settings because, for this domain, the key terms could be
the measures of the ontology view of Microstructural analy- related by formal relations with other concepts that do not
sis, considering the terms of Diagenesis. Notice that it is ex- belong to the Microstructural analysis ontology view.
pected that the resulting ratios are greater than 1. In a similar
way, the row Case 2, the table presents, for both approaches Table 2: Evaluation results. In this Talbe, P, R and F mean
A1 and A2, the ratios between the measures of the ontol- Precision, Recall and F-measure, respectively
ogy view for Microstructural analysis and the measures of
the ontology view of Diagenesis, considering the terms of Case Measure
Approach
A2
Microstructural analysis. A1
wP
wP wRT wP
- wP wFR wRT wRT
wRT wFR wFR
wFR
Evaluation of the Ontology View for Diagenesis Case
P 0.69 22.0 22.0 15.67 22.0 22.0 15.67 15.67 15.67
R 1.15 3.88 4.43 3.20 1.94 2.21 1.68 3.56 1.88
1
In the approach A1 (basic approach), the ontology view gen- F 0.97 9.25 9.0 7.60 9.25 9.0 7.60 7.60 7.60
p 2.39 0.75 0.84 0.72 0.54 0.64 0.52 0.81 0.62
erated for the Diagenesis community obtained a ratio of f- Case
R 1.46 3.38 3.69 3.15 5.23 5.62 4.77 3.46 5.15
2
measure smaller than 1. This means that the basic approach F 1.61 1.56 1.71 1.47 1.38 1.59 1.29 1.56 1.44
does not satisfy the expectation. This happens because the
approach A1 includes many ontology elements that are not
necessary for the community. However, the ratio of recall Conclusion
is greater than 1, as expected. This means that the ontology
view generated for the Diagenesis community contains more In this work, we propose a novel approach for extracting
relevant terms than the ontology views obtained for the Mi- well-founded ontology views, by improving the approach
crostructural analysis community. proposed in (Lozano et al. 2014). This approach eliminates
For the approach A2, proposed in this paper, in all cases, a source of information overload that is present in the pre-
the ratio of P , R and F was greater than 1. This occurs be- vious approach. Moreover, the proposed approach also pro-
cause it applies the principle of conservation of taxonomy vides more flexibility to the user, by allowing the control
just once to the target concepts, eliminating taxonomies that of important aspects of the process of extracting ontology
are useless for the community in focus. In general, approach views.
A2 results satisfy our expectations about the generated ontol- It is important to notice that, although this work adopts the
ogy view for all set of representative terms given in this case meta-properties defined by UFO, it can be viewed as a spe-
study. cific implementation of a more general idea. The notion of
In the column Case 1 of Table 2, it is possible to see that WFOV is defined according to a set of principles of conser-
the new approach (A2) achieve better results than the basic vation that should be followed by the extraction algorithm.
approach (A1), for the ontology views of the community of The set of principles of conservation can be changed, by
Diagenesis. The best results of approach A2 for Case 1 were including, excluding or modifying the principles proposed
achieved with wRT and with all the parameters as false. in this work. In this way, the general approach proposed in
(Lozano et al. 2014), and extended in this work, can be con-
sidered as independent of UFO.
Evaluation of the Ontology View for
In this work, we also propose a method for evaluating ap-
Microstructural Analysis proaches for extracting ontology views. This method uses
In Case 2, the ratio of f-measure achieved by the approach well-known measures used in information retrieval (preci-
A1 is greater than 1. Also, the precision obtained by this sion, recall and f-measure), for evaluating the fitness of the
approach achieved the highest value, in comparison with the resulting ontology views to the target conceptualization (of a
approach A2, considering all parameter combinations. community or task). According to this method, our novel ap-
The approach A2 also achieved high quality results in proach outperforms the basic approach (Lozano et al. 2014)
Case 2. Each combination of parameters of approach A2 sat- for extracting ontology views in most of the considered
isfies our expectations about the generated ontology views, cases. We hypothesize that this method can inspire methods
considering the set of representative terms given in this case that can be applied for evaluating ontology modules. This
study. However, the precision in all combination of param- hypothesis should be investigated in future works.
eters for Microstructural analysis ontology view has a ratio In future works, we also plan to improve the proposed
smaller than 1. This means that the ontology view gener- approach by identifying and eliminating other sources of in-
ated for the Microstructural analysis community contains formation overload in the resulting ontology views. Besides
few terms to cover the terminology used in the literature of that, we also intend to investigate if the ontological meta-
Microstructural analysis, than the ontology view for Diage- properties considered in this work can also be applied for
nesis. guiding the extraction of ontology modules.
References Seidenberg, J., and Rector, A. 2006. Web ontology segmen-
Abel, M.; Perrin, M.; and Carbonera, J. L. 2015. Ontological tation: Analysis, classification and use. In Proceedings of
analysis for information integration in geomodeling. Earth the 15th, WWW ’06, 13–22. New York, NY, USA: Interna-
Science Informatics 8(1):21–36. tional Conference on World Wide Web.
Abel, M. 2001. Study of Expertise in Sedimentary Petrogra- Worden, R., and Burley, S. 2003. Sandstone diagenesis:
phy and their importance for Knowledge Engineering. Ph.D. the evolution of sand to stone. Inernational Association of
Dissertation, Federal University of Rio Grande do Sul. Sedimentologists.
Bhatt, M.; Flahive, A.; Wouters, C.; Rahayu, W.; Taniar, D.;
and Dillon, T. 2004. A distributed approach to sub-ontology
extraction. In Proceedings..., volume 1, 636–641 Vol.1. Ad-
vanced Information Networking and Applications.
Carbonera, J. L.; Abel, M.; and Scherer, C. M. 2015. Vi-
sual interpretation of events in petroleum exploration: An
approach supported by well-founded ontologies. Expert Sys-
tems with Applications 42:27492763.
Carbonera, J. L.; Abel, M.; Scherer, C. M.; and Bernardes,
A. K. 2011. Reasoning over visual knowledge. In Joint
IV Seminar on Ontology Research in Brazil and VI Interna-
tional Workshop on Metamodels, Ontologies and Semantic
Technologies, 49–60.
Carbonera, J. L.; Abel, M.; Scherer, C. M.; and Bernardes,
A. K. 2013. Visual interpretation of events in petroleum
geology. In Proceedings of ICTAI 2013.
d’Aquin, M.; Sabou, M.; and Motta, E. 2006. Modulariza-
tion: a key for the dynamic selection of relevant knowledge
components. WoMo.
Doran, P.; Tamma, V.; and Iannone, L. 2007. Ontology mod-
ule extraction for ontology reuse: An ontology engineering
perspective. In Proceedings CIKM, 61–70.
Guizzardi, G. 2005. Ontological Foundations for Struc-
tural Conceptual Models. Ph.D. Dissertation, University of
Twente, The Netherlands.
Haertel, M., and Herwegh, M. 2014. Microfabric memory
of vein quartz for strain localization in detachment faults: A
case study on the simplon fault zone. Journal of Structural
Geology (0):–.
Lozano, J.; Carbonera, J. L.; Abel, M.; and Pimenta, M.
2014. Ontology view extraction: an approach based on on-
tological meta-properties. In Proceedings of ICTAI.
Lozano, J. 2014. Ontology view: a new sub-ontology ex-
traction method. Master’s thesis, Federal University of Rio
Grande do Sul.
Noy, N. F., and Musen, M. A. 2003. The prompt suite:
interactive tools for ontology merging and mapping. Inter-
national Journal of Human-Computer Studies 59(6):983 –
1024.
Noy, N., and Musen, M. 2009. Traversing ontologies to
extract views. In Stuckenschmidt, H.; Parent, C.; and Spac-
capietra, S., eds., Modular Ontologies, volume 5445 of Lec-
ture Notes in Computer Science. Springer Berlin Heidelberg.
245–260.
Powers, D. M. 2011. Evaluation: from precision, recall and
f-measure to roc, informedness, markedness and correlation.
International Journal of Machine Learning Technology 37–
63.