Observing the Impact and Adaptation to the Evolution of an Imported
Ontology
Omar Qawasmeh
1
, Maxime Lefranc¸ois
2
, Antoine Zimmermann
2
and Pierre Maret
1
1
Univ. Lyon, CNRS, Lab. Hubert Curien, UMR 5516, F-42023 Saint-
´
Etienne, France
2
Univ. Lyon, MINES Saint-
´
Etienne, CNRS, Lab. Hubert Curien, UMR 5516, F-42023 Saint-
´
Etienne, France
Keywords:
Ontology, Ontology Evolution, Ontology Adaptation.
Abstract:
Ontology evolution is the process of maintaining an ontology up to date with respect to the changes that arise
in the targeted domain or in the requirements. Inspired by this definition, we introduce two concepts related
to observe the impact and the adaptation to the evolution of an imported ontology. In the first one we target
the evolution of an imported ontology (if ontology O uses ontology O
0
, and then O
0
evolves). The second
one targets the adaptation to the evolution of the imported ontology. Based on our definition we provide a
systematic categorization of the different cases that can arise during the evolution of ontologies (e.g. a term t
is deleted from O
0
, but O continues to use it). We led an experiment to identify and count the occurrences of
the different cases among the ontologies referenced on two ontology portals: 1. the Linked Open Vocabulary
(LOV) ontology portal which references 648 different ontologies, 88 of them evolved. We identified 74 cases
that satisfy our definition, involving 28 different ontologies. 2. the BioPortal which references 770 different
ontologies, 485 of them evolved. We identified 14 cases that satisfy our definition, involving 10 different
ontologies. We present the observation results from this study and we show the number of different cases that
occurred during the evolution. We conclude by showing that knowledge engineers could take advantage of a
methodological framework based on our study for the maintenance of their ontologies.
1 INTRODUCTION
Ontologies play an important role in organizing and
categorizing data in information systems and on the
web. This leads to a better understanding, sharing and
analyzing of knowledge in a specific domain. How-
ever, domains are subject to changes, thus arises the
need to evolve ontologies in order to have an ad-
equate representation of the targeted domain (Sto-
janovic et al., 2003). Ontology evolution is the pro-
cess of maintaining an ontology up to date with re-
spect to the changes that might arise in the described
domain, and/or in the requirements (Zablith et al.,
2015).
Re-usability is considered as a good practice
while designing an ontology (Simperl, 2009). On the
one hand, re-usability saves time for knowledge engi-
neers while developing ontologies, but on the other
hand it raises the problem of adapting one’s ontol-
ogy to the evolution of an imported ontology and thus
complicates the maintenance process. In this paper
we are interested in observing the impact of the evo-
lution of an imported ontology, and the adaptation to
that evolution.
For example the ontology Schema.org core and
all extension vocabularies (O) has a version that was
published in 2014-10-30 (v
1
), and that uses terms
from the ontology Schema.org (O
0
) that was created in
2012-04-27 (v
0
1
). O
0
evolved to v
0
2
in 2017-03-23, and
similarly, O evolved to v
2
in 2017-05-19. The term
Bacteria was deleted in v
0
2
, yet v
2
still uses it. This
has an impact (O needs to be adapted), and illustrates
an issue that might arise when an imported ontology
evolves. We propose a frame for observing the impact
and adaptation to the evolution of an imported ontol-
ogy. Then we identify and try to explain occurrences
of such cases in the history of two ontology portals.
The rest of the paper is organized as follows: Sec-
tion 2 presents our motivating example. Section 3
presents an overview of ontology evolution. Section 4
proposes a theoretical analysis of the need for changes
that could stem from the evolution of an imported on-
tology. Section 5 presents an exhaustive theoretical
analysis of how an ontology may be adapted to such
evolution. Section 6 presents our method to analyze
the LOV portal and the BioPortal based on our defi-
76
Qawasmeh, O., Lefrançois, M., Zimmermann, A. and Maret, P.
Observing the Impact and Adaptation to the Evolution of an Imported Ontology.
DOI: 10.5220/0008064700760086
In Proceedings of the 11th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2019), pages 76-86
ISBN: 978-989-758-382-7
Copyright
c
 2019 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved

nition. Section 7 shows the evaluation of our experi-
ments. Section 8 discusses our results. Finally, Sec-
tion 9 concludes the paper.
2 MOTIVATING EXAMPLE
Amal is a knowledge engineer who works on de-
veloping an ontology to describe the child care do-
main. Let’s assume that she created an ontology
called Childcare. In the version V1.1 of Childcare,
created in May 2017, Amal used a specific term
programmOfStudy from another ontology called Ed-
ucation created in January 2017. Childcare contains
at least a link to a term from Education. In September
2017 the creators of the Education ontology released
the version V1.2. Amal does not notice the evolu-
tion. Thus, she thinks that her ontology is still using
the V1.1 version of the Education ontology. Several
issues might arise:
1. The term programmOfStudy was removed from
Education, however it is still used Childcare. This
might have an impact over Childcare. As a con-
sequence for this impact, Amal should adopt the
changes and change her ontology accordingly.
2. New terms were introduced in Education,V1.2
(e.g. boarding school). Amal should be made
aware of these new terms in order to possibly
make use of them in her ontology.
3 AN OVERVIEW ON
ONTOLOGY EVOLUTION
In this section we present the life cycle of ontology
evolution as defined by (Zablith et al., 2015), and
some related works. The life cycle (Zablith et al.,
2015) consists of five phases summarized as: 1. De-
tect the need for evolution by either studying the
users’ behavior while using ontology-based systems
or by analyzing the data sources that use the ontology,
2. Suggest changes to evolve the ontology, 3. Vali-
date the suggested changes before adopting them into
the ontology, 4. Assess and study the impact of the
evolution, by evaluating the impact of the changes on
external artifacts that rely on the ontology (e.g. other
ontologies, systems) and/or the cost of performing the
changes. 5. Keep track of the implementation of the
changes, in order to facilitate the management of the
different versions that are created during the evolu-
tion process. In this research, we focus on both the
first and fourth phases.
Authors in (Zablith et al., 2015) mentioned several
research papers that tackle the problem of detecting
the need of evolution. In addition, two papers (Tartir
et al., 2010, Papavassiliou et al., 2009) are highly rel-
evant for our approach that will rely on the structural
changes to detect the need of the evolution (i.e. detec-
tion of addition and deletion of terms). In (Tartir et al.,
2010), the authors emphasize what came from (Noy
and Musen, 2002) and they mentioned that ontology
evolution is caused by three reasons: 1. Changes in
the described domain, 2. Changes in the conceptual-
ization (e.g. deletion and addition), and 3. Changes
in the explicit specification. In (Papavassiliou et al.,
2009) a change detection algorithm is proposed and
relies on a specific language they also proposed. One
feature of their algorithm is to detect the need of evo-
lution out of the changes that happen, such as renam-
ing a class (i.e. delete and add).
Several approaches have studied the impact of on-
tology evolution in different scenarios. Dragoni and
Ghidini (Dragoni and Ghidini, 2012) investigate how
ontology evolution operations affect the effectiveness
of search systems. They focused on three operations:
1. rename a concept, 2. delete a concept, and 3. move
a concept. They analyzed the impact of the evolution
of the ontology over a search system: they performed
75 queries over a search system at every version of
the evolved ontology and they calculated the effec-
tiveness of the system by comparing with a baseline.
Abgaz et al. (Abgaz et al., 2012) analyzed struc-
tural impact and semantic impact over ontologies.
They defined a set of rules to analyze the impact
by detecting unsatisfiable statements and wrong in-
stances. They defined 10 change operations that cover
the different change scenarios.
Groß et al. (Groß et al., 2012) investigated how
the changes in the Gene ontology
1
might affect the
statistical applications for the experimental and simu-
lated data (external artifacts). CODEX tool (Hartung
et al., 2012) was used to detect the changes (e.g. ad-
dition, merging, moving). They introduced their own
stability measure by choosing a fixed set of genes to
compute the experimental result set at different point
of time with freely chosen ontology and annotation
versions.
Mihindukulasooriya et al. (Mihindukulasooriya
et al., 2016) introduced a study that shows how DBpe-
dia (Lehmann et al., 2015), Schema.org (Guha et al.,
2016), PROV-O (Lebo et al., 2013) and FOAF (Brick-
ley and Miller, 2010) ontologies evolved through their
life time. They counted the changes that occurred
between the different versions such as, addition and
deletion of classes, properties, sub-classes and sub-
1
http://geneontology.org/
Observing the Impact and Adaptation to the Evolution of an Imported Ontology
77

properties. They show that ontology evolution is more
challenging when the ontology size is large. More-
over, they show the need of having tools that can help
during the evolution process.
Abdel-Qader et al. (Abdel-Qader et al., 2018) ana-
lyzed the impact of the evolution of terms in 18 differ-
ent ontologies referenced in LOV. Their method con-
sisted of two phases: 1. retrieve all the ontologies that
have more than one version, and 2. investigate how
terms are changed and adopted in the evolving ontolo-
gies. They applied their analysis on three large-scale
knowledge graphs: DyLDO
2
, BTC
3
and Wikidata.
4
They found that some of the term changes in the 18
ontologies are not mapped into the three knowledge
graphs. Also they concluded that there is a need for a
service to keep an eye on the ontology changes. They
claim that it would help the knowledge engineers and
the data publishers maintaining their artifacts (other
ontologies, systems or data sets).
As a conclusion for this study, we show that these
different approaches studied the impact of ontology
evolution by three techniques: 1. Observing the struc-
tural changes such as: addition, deletion, and moving
(i.e. deletion then addition) during the life time of an
ontology. 2. Measuring the impact of the evolution
over the different artifacts (e.g. search systems) that
might rely on the evolved ontologies. 3. Listing the
changes, the frequency of each change, and the time
it took to adopt a specific set of changes. In this pa-
per, we follow the first technique, and we propose an
approach that observes the evolution of terms of an
imported ontology.
4 OBSERVING THE IMPACT OF
THE EVOLUTION OF AN
IMPORTED ONTOLOGY
As shown in the previous section, there is a need to
detect when to perform changes on ontologies. This
is defined in the first phase of the ontology evolution
life-cycle. In this research we are interested in observ-
ing the impact of the evolution of an imported ontol-
ogy, and the adaptation to that evolution. We observe
the evolution of terms that have the namespace of the
imported ontology. A RDF term is generally defined
as: I ∪ B ∪ L
5
. In this research we take into consider-
ation only the IRIs (I).
2
http://km.aifb.kit.edu/projects/dyldo/data
3
https://km.aifb.kit.edu/projects/btc-2012
4
https://www.wikidata.org
5
I: IRIs, B: blank nodes, and L: literals
v1
Ngeo ontology
DCE ontology
v1` v2`
Uses
14-06-2012
11-10-2010
05-02-2012
Creation time
Creation time
Figure 1: A time line showing the creation times of the
Ngeo ontology and the DCE ontology, where Ngeo uses
terms that are defined by DCE.
Detecting the need for evolving the set of terms
can be manifested through two behaviors:
1. There is already a problem: If an ontology O uses
a term t that has the namespace of another ontol-
ogy O
0
, however it is not defined in O
0
.
2. A problem has occurred because of the evolution
process: Let’s assume that there is an ontology O
that uses a term t that has the namespace of an-
other ontology O
0
. O
0
evolved which cause the
deletion of t. This evolution might cause prob-
lems for O. This raises the need to evolve O in
order to reflect the new changes, which leads to
solve the different problems.
We give our definition of an imported ontology
evolution hereafter:
Definition 1. Imported Ontology Evolution. Im-
ported ontology evolution is a situation where: O is
an ontology which has at least one version v
1
. O
0
is a
different ontology which has at least two versions v
0
1
and v
0
2
. O uses terms that have the namespace of O
0
.
time(v) is the creation time for a version.
A case of imported ontology evolution is noted
hv
1
, v
0
1
, v
0
2
i and holds when the following conditions
are satisfied:
time(v
0
1
) < time(v
0
2
) ∧ time(v
0
1
) < time(v
1
)
Fig 1 presents an example of one case of imported
ontology evolution: the NeoGeo Geometry Ontol-
ogy has one version (v
1
: ngeo 2012-02-05) that uses
the Dublin Core Metadata Element Set ontology (v
0
1
:
dce 2010-10-11, v
0
2
: dce 2012-06-14). Table 1 re-
flects the different cases that may occur with respect
to Definition 1. t is a term that has the namespace of
O
0
. The circles at every line represent the set of terms
(t) that exist in the the two versions of the ontology
O
0
(i.e. v
0
1
and v
0
2
). The columns represents the set
of terms t that exist in the ontology O (i.e. v
1
). Four
possible cases might happen:
Row 1. No changes over t
Case 1.a There is no change of t to detect, therefore
there is no interest in studying this case.
This case holds for all the terms t with the
namespace of O
0
, that are neither defined in
O
0
nor used in O
KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development
78

Case 1.b This case holds when O uses a term t with
the namespace of O
0
, but that is not defined
in O
0
. Some terms that have the namespace
of O
0
are being used in v
1
without being de-
fined before. This is a mistake, hence there
is a need to evolve v
1
to reflect the latest
changes.
Row 2. t is deleted in v
0
2
The owners of O
0
decided to stop using a term (e.g.
programmOfStudy) in v
0
2
:
Case 2.a The term is not used in v
1
. No problems to
be reported, and v
1
was not affected by the
evolution of O
0
.
Case 2.b During the evolution, the term t was
deleted. However, it is still being used in
v
1
. This might introduce a problem of us-
ing terms that does not exist anymore. So v
1
should evolve to better reflect the changes
of O
0
.
Row 3. t exist in both v
0
1
and v
0
2
There is no changes on t:
Case 3.a The term is not used in v
1
. However, it can
be recommended for use in the upcoming
versions of v
1
.
Case 3.b No changes over the terms during the evo-
lution.This case is not problematic.
Row 4. t is added to v
0
2
The owners of O
0
introduced a new term (e.g.
boardingSchool) in v
0
2
:
Case 4.a The term t is not used in v
1
. It can be inter-
esting to use, thus this addition can be noti-
fied.
Case 4.b The term t is used in v
1
, however it was de-
fined later in v
0
2
.
Table 1: The set of cases that might happen during the evo-
lution of O
0
considering a term t that has the namespace of
O
0
.
Case 1.a: No changes
occurred
Case 1.b: Term is used in v
1
without being defined in O`
Case 2.a: No impact
occurred
Case 2.b: There is a need
for evolution, because the
term is no longer in O`
Case 3.a: No impact
occurred. Suggest to add
new terms
Case 3.b: No impact
occurred
Case 4.a: Suggest to add
new terms
Case 4.b: Term used before
it is defined
O`
O
v
1
`
1
2
3
4
a
b
v
2
`
v
1
uses
v
1
v
1
`
v
2
`
v
1
`
v
2
`
v
1
`
v
2
`
v1 v2
v1` v2`
14-06-2011
20-04-2010
22-07-2013
28-11-2010
Figure 2: A time line showing the creation times of the mu-
sic ontology (mo) and the bio ontology (bio), where mo uses
terms that are defined by bio.
5 OBSERVING THE
ADAPTATION TO THE
EVOLUTION OF AN
IMPORTED ONTOLOGY
We have seen the impact of the evolution of an im-
ported ontology. As a consequence of this evolu-
tion, the impacted ontology should be adapted to the
changes and evolve accordingly. This creates a situa-
tion which we call ontology co-evolution.
The term “ontology co-evolution” has been al-
ready used in three research papers. Authors in
(Kupfer et al., 2006, Kupfer and Eckstein, 2006) de-
fine the co-evolution as the integration between the
database schemas and ontologies to design and evolve
the targeted ontologies. Also (Ottens et al., 2007) de-
fines the co-evolution as the creation of ontologies by
taking advantage of natural language techniques to
process some raw text. These definitions are irrele-
vant to the problem we investigate. We define ontol-
ogy co-evolution as:
Definition 2. Ontology Co-Evolution. Ontology co-
evolution is a situation where: O is an ontology which
has at least two versions v
1
and v
2
. O
0
is a different
ontology which has at least two versions v
0
1
and v
0
2
. O
uses terms that have the namespace of O
0
. time(v) is
the creation time for a version.
In order to have a co-evolution case between
O and O
0
with the ontologies hv
1
, v
0
1
, v
2
, v
0
2
i, the
following condition must be satisfied:
time(v
1
) < time(v
2
) ∧ time(v
0
1
) < time(v
0
2
) ∧
time(v
0
1
) < time(v
1
) ∧ time(v
0
2
) < time(v
2
)
As mentioned in the previous section, in this re-
search we take into consideration only the IRIs (I).
We provide an exhaustive categorization of the differ-
ent cases that can arise during the adaptation to on-
tology evolution (i.e. co-evolution). Fig 2 presents
an example of one case of ontology co-evolution: the
Music ontology has two versions (v
1
: mo 2010-11-
28, v
2
: mo 2013-07-22) that are respectively using
the two versions of the Bio ontology (v
0
1
: bio 2010-
04-20, v
0
2
: bio 2011-06-14).
During the evolution of O
0
, terms may be intro-
duced or deleted. We identified the occurrences of
Observing the Impact and Adaptation to the Evolution of an Imported Ontology
79

adaptation to ontology evolution of O and O
0
. We ob-
serve the set of terms that have the namespace of O
0
.
Table 2 shows the different cases that may occur. The
left circles represent the set of terms that exist in the
first version of an ontology (i.e. v
1
and v
0
1
), and the
right circles represent the set of terms that exist in the
second version of an ontology (i.e. v
2
and v
0
2
). t is a
term that has the namespace of O
0
.
In our illustrating example of Section 2, let us as-
sume that Amal finally noticed the evolution of Ed-
ucation ontology and decided to evolve her ontol-
ogy Childcare to V1.2 on November 2017. Based
on our definition, the ontology Childcare is consid-
ered as O which has two versions v
1
: Childcare V1.1,
created in May 2017 and v
2
: Childcare V1.2, cre-
ated in November 2017. The ontology Education is
considered as O
0
and has two versions v
0
1
: Education
V1.1, created in January 2017 and v
0
2
: Education V1.2,
created in September 2017. Amal is using the term
programmOfStudy from O
0
. Following each line of
Table 2, the following set of cases might occur during
the life journey of Amal’s ontology:
Row 1. No changes over the terms of v
0
1
, or v
0
2
Case 1.a There is no change of t to detect, therefore
there is no interest in studying this case.
Case 1.b Amal made a typo by using the term
programOfStudy (i.e. program is written
with one “m” instead of two) in v
1
, but then
she realizes that this term does not exist in
O
0
. She fixes this mistake by not using it in
v
2
anymore.
Case 1.c Amal uses t in both v
1
and v
2
. This case
might be explained by the fact that t is de-
fined in a previous version (e.g. v
0
) of the
ontology O
0
(i.e. t(v
0
0
) < t(v
0
1
)).
Case 1.d Amal introduces a mistake by using t in v
2
.
Row 2. t is deleted in v
0
2
The owners of O
0
decided to stop using the term
programmOfStudy in v
0
2
:
Case 2.a Amal does not use t that was recently
deleted. Hence v
1
and v
2
were not affected.
Case 2.b Amal realizes that t was deleted, so she
stops using it in v
2
.
Case 2.c Amal does not realize the deletion of t, and
she keeps using it in v
2
.
Case 2.d Amal starts to use t in her second version
(v
2
), which introduces a mistake.
Row 3. t exist in both v
0
1
and v
0
2
None of the cases (3.a, 3.b, 3.c, and 3.d) is problem-
atic.
Row 4. t is added to v
0
2
The owners of O
0
introduced a new term
boardingSchool in v
0
2
:
Case 4.a Amal has not noticed the addition of t, even
if it might be interesting for her to introduce
it.
Case 4.b Amal was already using t in (v
1
), but she
decided to remove it from v
2
.
Case 4.c Amal was already using t in v
1
, and she con-
tinues using it in v
2
.
Case 4.d Amal realizes the addition of t, and she start
using it in v
2
.
Cases 4.b and 4.c are corner cases that are discussed
further in section Section 8.
6 ANALYZING THE
ADAPTATION OF THE
EVOLUTION OVER
ONTOLOGY PORTALS
The main aim of ontology portals is to group dif-
ferent ontologies in order to facilitate the process of
finding and reusing them. Moreover, ontology por-
tals are convenient to group the different versions of
each ontology. These versions are time stamped in
order to keep track of the creation time for each one.
This facilitates our process to extract the time of each
version. This section presents the two ontology por-
tals we used and the method we applied to detect the
occurrences of ontology co-evolution. Section 6.1
presents LOV portal (Vandenbussche et al., 2017),
and Section 6.2 presents BioPortal (Whetzel et al.,
2011). We have selected these two portals because
they are the ones that reference the greatest number
of vocabularies available on the Web.
6.1 Analyzing the Adaptation of
Ontology Evolution over the Linked
Open Vocabulary (LOV)
Linked Open Vocabulary (LOV) (Vandenbussche
et al., 2017) is considered a rich repository of ontolo-
gies. LOV’s main goal is to help publishers and users
of linked data and vocabularies to assess, reuse, and
publish different vocabularies based on their needs.
LOV currently references 648 different vocabularies
6
,
each one being described with different properties,
6
Last counted on June 2018
KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development
80

Table 2: The set of cases that might happen during the ontology co-evolution considering t that has the namespace of O
0
.
Case 1.a: No
changes occurred
Case 1.b: Term is
used in v
1
but doesn’t
exist in O`
Case 1.c: Term is
used in both v
1
and v
2
but doesn’t exist O`
Case 1.d: Term is
used in v
2
but doesn’t
exist O`
Case 2.a: Term is
deleted from v
2
` and
not used in O
Case 2.b: Term is
deleted in v
2
` but still
used in v
1
Case 2.c: Term is
deleted in v
2
` but still
used in both v
1
and v
2
Case 2.d: Term is
deleted in v
2
` and still
used in v
2
Case 3.a: Term exists
in both v
1
` and v
2
` and
not used in O
Case 3.b: Term exists
in both v
1
` and v
2
` and
used in v
1
Case 3.c: Term exists
in both v
1
`
and v
2
`
and
used in both v
1
and v
2
Case 3.d: Term exists
in both v
1
` and v
2
` and
used in v
2
Case 4.a: Term
introduced in v
2
` and
not used in O
Case 4.b: Term exists
in v
2
` and used in v
1
Case 4.c: Term exists
in v
2
` and used in both
v
1
and v
2
Case 4.d: Term exists
in v
2
` and used in v
2
O`
O
v
1
`
1
2
3
4
a b c d
v
2
`
v
1
v
2
uses
v
1
`
v
2
`
v
1
`
v
2
`
v
1
`
v
2
`
v
1
v
2
v
1
v
2
v
1
v
2
such as number of incoming links (i.e. how many on-
tologies are using ontology O), number of outgoing
links (i.e. how many ontologies are used by ontology
O), number of different versions, and datasets that are
using ontology O.
Out of the 648 ontologies of LOV, there are 88 on-
tologies that have evolved with a total number of 344
versions. The number of different versions that is as-
sociated with each ontology varies. For example the
FOAF
7
ontology has 10 available versions to down-
load from the LOV knowledge base. Fig 3 shows the
relation between the total number of versions and the
number of ontologies that have the specific number of
versions.
Number of versions
Number of Ontologies
0
25
50
75
100
1 5 10 50 100
LOV BioPortal
Number of Ontologies vs. Number of versions
Figure 3: The relation between the total number of versions
and the number of ontologies that have the specific number
of versions for the ontologies that are referenced in LOV
and BioPortal.
7
Available at: https://lov.linkeddata.es/dataset/lov/vocab
s/foaf
However, not all the ontologies that evolved are
connected (i.e. using terms from each other). In order
to retrieve the set of ontologies that satisfy the condi-
tions in Definition 2, we first issued a SPARQL query
(Listing 1.1) over the LOV RDF dump in order to re-
trieve all ontologies that have at least two different
versions and at least 1 incoming link. The result is 46
different ontologies and a total of 205 different ver-
sions.
Second, we used Apache-Jena in order to get all
the different ontologies that have more than one out-
going links. This decreased the list of versions to 198.
Third, we extracted all the creation times for the dif-
ferent ontologies versions, and we filtered them based
on the selection criteria of Definition 2. As a result
we identified 74 cases of ontology co-evolution, in-
volving 28 different ontologies.
6.2 Analyzing the Adaptation of
Ontology Evolution over BioPortal
BioPortal (Whetzel et al., 2011) is an open repository
for biomedical ontologies. BioPortal’s main goal is to
make biomedical knowledge and data available on the
internet using ontologies. This is useful for boosting
biomedical science and clinical care domains. Bio-
Portal currently references 770 different ontologies,
8
each one being described with different properties,
such as the number of different versions, along with
general metrics (e.g. number of classes, properties
8
Last counted on January 2019
Observing the Impact and Adaptation to the Evolution of an Imported Ontology
81

PREFIX voaf:<http://purl.org/vocommons/voaf#>
PREFIX dcterms: <http://purl.org/dc/terms/>
PREFIX dcat: <http://www.w3.org/ns/dcat#>
SELECT DISTINCT ∗ WHERE {
GRAPH <http://lov.okfn.org/dataset/lov>{
?vocab a voaf:Vocabulary ;
dcterms:title ?title ;
dcterms:modified ?modified ;
voaf:reusedByDatasets ?dataset ;
voaf:reusedByVocabularies ?import ;
dcat:distribution ?distrib .
?distrib dcterms:issued ?issued .
FILTER ( !isBlank(?distrib) )
FILTER ( ?import >0 )
BIND (STR(?issued) AS ?date)
}
} ORDER BY DESC(?dataset)
Listing 1: This query returns all ontologies which have at
least 2 versions and at least 1 incoming link.
and instances).
Out of the 770 ontologies of BioPortal, 485 on-
tologies have evolved with a total number of 15025
versions. The number of different versions that is as-
sociated with each ontology varies. For example the
HIV
9
ontology has 12 available versions. Fig 3 shows
the relation between the total number of versions and
the number of ontologies that are associated with each
ontology. As explained earlier, only some of the on-
tologies that evolved satisfy our configuration of on-
tology co-evolution.
In order to retrieve the set of ontologies that sat-
isfy the conditions in Definition 2, we firstly used the
BioPortal API
10
to retrieve all ontologies that have
at least two versions. As a result, we got 485 on-
tologies that have 15025 different versions. Secondly,
we used Apache-Jena in order to get all different on-
tologies that have more than one outgoing link. And
thirdly, we extracted all the creation times for the dif-
ferent ontology versions, and we filtered them based
on the selection criteria of Definition 2. We identi-
fied 14 cases of ontology co-evolution, involving 10
different ontologies.
9
Available at: http://data.bioontology.org/ontologies/
HIV/submissions
10
http://data.bioontology.org/documentation
7 IDENTIFICATION OF THE
OCCURRENCES OF
ADAPTATION TO ONTOLOGY
EVOLUTION
In this section, we present the results of an exper-
iment
11
to detect ontology co-evolution using the
cases that are defined in Section 5. In Section 8 we
discuss these results in more details.
We retrieved from LOV a set of 28 ontologies with
74 co-evolution instances. As for BioPortal we re-
trieved a set of 10 ontologies with 14 co-evolution in-
stances.
12
We extracted the set of terms for each version,
and the name spaces for the used ontologies (O
0
’s ver-
sions), and we used them to compute the number of
occurrences of the different cases.
Table 3 shows the number of occurrences for each
co-evolution case for LOV (first value in each cell)
and BioPortal (second value).
Table 3: The number of occurrences for each co-evolution
case for LOV (first value) and BioPortal (second value) with
respect to namespace of (O
0
).
0
130
3
929
3
115
23
27
0
0
0
3
0
0
16875
9135
10
0
270
2058
23
0
2420
1560
0
115
0
908
0
0
O`
O
v
1
`
1
2
3
4
a b c d
v
2
`
v
1
`
v
2
`
v
1
`
v
2
`
v
1
`
v
2
`
v
1
v
2
v
1
v
2
v
1
v
2
v
1
v
2
uses
Now we discuss further different interesting cases.
In the following subset of cases, a version of O uses
a term t with the namespace of O
0
, however t is not
defined in the two versions of O
0
:
Case 1.b (i.e. a term t is used in v
1
, however
it does not exist in v
0
1
and v
0
2
): This co-evolution
case occurred 130 times in BioPortal but never in
LOV. An example is the co-evolution process of the
Schema.org core and all extension vocabularies (v
1
:
created in 2014-10-30 and v
2
: created in 2017-05-19),
with Schema.org ontology (v
0
1
: created in 2012-04-27
and v
0
2
: created in 2017-03-23). The terms Bacteria,
11
Can be found at: https://github.com/OmarAlqawasmeh/
coEvolutionTermsExtraction (Full results can be found
inside resources folder)
12
The co-evolution cases of LOV and BioPortal are
inside the resources folder at https://github.com/ Omar-
Alqawasmeh/coEvolutionTermsExtraction
KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development
82

FDAcategoryC and Diagnostic are used in v
1
, how-
ever they do not exist in v
0
1
and v
0
2
.
Case 1.c (i.e. a term t is used in both v
1
and v
2
,
however it does not exist in v
0
1
and v
0
2
): This case oc-
curred 3 times in LOV and 929 times in BioPortal.
An example is the co-evolution process of the Sta-
tistical Core Vocabulary (v
1
: created in 2011-08-05
and v
2
: created in 2012-08-09), with DCMI Metadata
Terms (v
0
1
: created in 2010-10-11 and v
0
2
: created in
2012-06-14). The terms dc:status and dc:partOf
are used in v
1
and v
2
, however they do not exist in v
0
1
and v
0
2
.
Case 1.d (i.e. a term t is used in v
2
, however
it does not exist in v
0
1
and v
0
2
): This case occurred 3
times in LOV and 115 times in BioPortal. An ex-
ample is the co-evolution process of the Europeana
Data Model vocabulary (v
1
: created in 2012-01-23
and v
2
: created in 2013-05-20), with Dublin Core
Metadata Element Set (v
0
1
: created in 2010-10-11 and
v
0
2
: created in 2012-06-14). The terms dc:issued
and dc:modified are used in v
2
however they do not
exist in v
0
1
and v
0
2
.
In the following subset of cases a term t is defined
in v
0
1
and is deleted in v
0
2
:
Case 2.a (i.e. a term t is deleted in v
0
2
, and it is
not used in any of O’s versions): This case occurred
23 times in LOV and 27 times in BioPortal. This is
a normal case, and no problem occurred during the
co-evolution.
Case 2.b (i.e. a term t is deleted in v
0
2
, and then
deleted in v
2
): This case has no occurrences in LOV
nor in BioPortal. We are not discussing it further.
Case 2.c (i.e. a term t is deleted in v
0
2
, however it is
used in v
1
and still in v
2
): This case occurred 3 times
in BioPortal. It shows a problem of using terms that
do not exist anymore in O
0
. For example in the co-
evolution process of the Schema.org core and all ex-
tension vocabularies (v
1
: created in 2014-10-30 and
v
2
: created in 2017-05-19), with Schema.org ontology
(v
0
1
: created in 2012-04-27 and v
0
2
: created in 2017-
03-23). The terms MedicalClinic, Optician and
VeterinaryCare are used in both v
1
and v
2
, however
they do not exist in the latest version of O
0
(these dif-
ferent terms were deleted from v
2
of Schema.org).
Case 2.d (i.e. a term t is deleted in v
0
2
, however it
is added in v
2
): This case has no occurrences in LOV
nor in BioPortal. We are not discussing it further.
Cases 3.a, 3.b, 3.c, and 3.d are not problematic
cases, so they are not investigated further.
In the following subset of cases a term t is intro-
duced in v
0
1
:
Case 4.a (i.e. a term t is added in v
0
2
, and it was not
used in v
1
or v
2
): There were 2420 terms added to v
0
2
in the ontologies that are referenced in LOV, and 1560
terms were added to v
0
2
in the ontologies that are refer-
enced in BioPortal. These different terms are not used
in v
1
or v
2
. An example is the co-evolution process of
the Semanticscience Integrated Ontology (SIO) (v
1
:
created in 2015-06-24 and v
2
: created in 2015-09-02),
with The Citation Typing Ontology (CITO) (v
0
1
: cre-
ated in 2010-03-26 and v
0
2
: created in 2015-07-03).
The term isDocumentedBy from O
0
could be useful
to use by O, thus the owners of O can be notified and
recommended to use it.
Case 4.b (i.e. a term t is added in v
0
2
, and it was
already used in v
1
): This case occurred 115 times in
BioPortal, and it has no occurrence in LOV. An ex-
ample is the co-evolution process of the Schema.org
core and all extension vocabularies (v
1
: published
in 2014-10-30 and v
2
: published in 2017-05-19),
with Schema.org ontology (v
0
1
: created in 2012-04-
27 and v
0
2
: created in 2017-03-23). The terms
SoundtrackAlbum, Hardcover and SingleRelease
are used in v
1
, however they were introduced later in
v
0
2
.
Case 4.c (i.e. a term t is added in v
0
2
, and it was
already used in both of O’s versions): This case oc-
curred 951 times in BioPortal, and it has no occur-
rence in LOV. An example is the co-evolution process
of the Semanticscience Integrated Ontology (SIO) (v
1
:
created in 2015-06-24 and v
2
: created in 2015-09-02),
with The Citation Typing Ontology (CITO) (v
0
1
: cre-
ated in 2010-03-26 and v
0
2
: created in 2015-07-03).
The term citesAsAuthority is used in both v
1
and
v
2
, however it was introduced in v
0
2
.
Case 4.d (i.e. a term t is added in v
0
2
, and v
2
starts
to use it): This case has no occurrences in LOV or
BioPortal, so we are not investigating them.
8 DISCUSSION
Our approach relies on the described situation in Def-
inition 2. However we can stress some comparison
with (Abdel-Qader et al., 2018):
1. Both analyses observe the additions and dele-
tions of terms, however in (Abdel-Qader et al.,
2018) they observe how the terms are changed and
adopted in the evolving ontologies, where we ob-
serve the changes when two ontologies are con-
nected to each other.
2. 13 ontologies with 37 versions that are referenced
in LOV were used in the experiments of (Abdel-
Qader et al., 2018), where we retrieved a total of
38 ontologies referenced in LOV and BioPortal
and we observed 88 evolution cases.
Observing the Impact and Adaptation to the Evolution of an Imported Ontology
83

After analyzing the results, we confirm the ob-
servation of (Groß et al., 2012, Kirsten et al., 2009)
showing that in general the addition of terms occurs
more frequently than the deletion of terms during the
evolution process. Table 4 shows the number of added
terms comparing to the number of deleted terms for
the set of ontologies that are referenced in LOV and
BioPortal and satisfy our definition of co-evolution.
Table 4: Number of added terms comparing to number of
deleted terms in both LOV and BioPortal.
Portal LOV BioPortal
Added Deleted Added Deleted
v
2
26 10 115 245
v
0
2
2420 23 2583 30
Our results can be conveniently presented in three
categories that clearly appear in our experiment:
1. Assessment of Good Practices 2. Detection of
Wrong Practices 3. Uncertain cases. In the next sub-
sections we discuss these different categories.
8.1 Assessment of Good Practices
Case 1.b in BioPortal shows a good practice from the
owners of O. They noticed that the term t is not de-
fined in both v
0
1
and v
0
2
, so they decided to delete it
from v
2
.
Case 2.a in LOV and Bioportal show a good prac-
tice from the owners of O
0
. They noticed that the term
t is not used in both v
1
and v
2
so they decided to delete
it from v
0
2
.
Cases 2.b and 2.d have no occurrences in all of
the ontologies that are referenced in both LOV and
BioPortal. This is the normal case of ontology evo-
lution, and it is one indicator of the quality of the
co-evolution. For instance, case 2.b indicates that the
set of ontologies stops using the terms after they have
been deleted in O
0
, and case 2.d indicates that there
were no mistake of using the set of deleted terms in
the newest version of O.
8.2 Detection of Wrong Practices
Cases (1.c and 1.d) from LOV and cases (1.c and
1.d) from BioPortal, demonstrate the problem of us-
ing terms that do not exist in v
0
1
and v
0
2
. A possible
explanation is that these terms were used from a pre-
vious version of O
0
. Let’s assume that this previous
version is v
0
0
, then these cases can happen only if the
publishing time of t(v
0
0
) is before the publishing time
of t(v
0
1
). In these cases, the owners of O, should be
notified of the changes, and they should be suggested
to delete the terms that do not exist anymore.
Case 2.c from BioPortal shows that some terms
are still used in both of O’s versions after being
deleted from O
0
. In order to prevent such kind of
problems the owners should be notified about these
cases.
Case 4.b from BioPortal shows that some terms
have been already used in v
1
, however they were
added later in v
0
2
. The v
1
of Schema.org core and all
extension vocabularies uses terms that were later de-
fined by v
0
2
of Schema.org ontology. The Schema.org
core and all extension vocabularies is an extension of
Schema.org, however it has its own namespace. Each
reviewed extension for schema.org has its own chunk
of schema.org namespace (e.g. if extension name is
x, the namespace of this extension is x1.schema.org).
13
We retrieved all terms that have the namespace
of Schema.org.
14
Other terms with different names-
paces were discarded.
15
This reflects a bad practice
in a way of using terms that have not been defined in
the second version. These terms could be harbinger
to add in the next versions.
Case 4.c from BioPortal shows that some terms
are used in v
1
and v
2
, however they were firstly intro-
duced in v
0
2
. Both of v
1
and v
2
of Semanticscience In-
tegrated Ontology (SIO) use terms that were later de-
fined by v
0
2
of The Citation Typing Ontology (CITO).
The term citesAsAuthority was firstly defined in
v
0
2
The Citation Typing Ontology (CITO), however
there is an object property that has the same name
in v
0
1
. One explanation for this kind of errors is that
the knowledge engineers might introduce a typos dur-
ing the development process of the ontology. In these
cases, the owners of O, should be notified that the
term they use is not a term. They should look at it
carefully and possibly delete it.
8.3 Uncertain Cases
In cases (3.a, 3.b, 3.c and 3.d) from LOV and Biopor-
tal, there was no change of terms in the two versions
of O
0
. This indicates that the co-evolution process has
no problem to report. Some terms are shared between
v
0
1
and v
0
2
so there was no addition or deletion over
them.
Cases 4.a and 4.d in both LOV and BioPortal show
the number of terms that were added during the evo-
lution of O
0
. These terms were not used in any of O’s
13
More details about the extensions man-
aging of schema.org can be found at:
https://schema.org/docs/extension.html
14
Namespace of Schema.org is http://schema.org/
15
Some examples of discarded namespaces:
https://health-lifesci.schema.org/, https://pending.schema.
org/, https://meta.schema.org/
KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development
84

versions. These cases can be explained in two ways:
1. The owners of O did not notice the addition of
these terms, however they might be interested in
using some of these new terms. This might in-
troduce a problem, thus further content analy-
sis should be introduced to possibly recommend
changes to the owners.
2. The owners of O noticed the addition of these
terms and they decided not to add them.
9 CONCLUSION AND FUTURE
WORK
As shown from the literature review, there is a need to
formalize a conceptual frame for assessing the impact
of ontology evolution and for tackling the different is-
sues that arise during the evolution of ontologies. In
this paper we presented a situation of ontology evo-
lution which considers the evolution of an ontology
O that imports another one O
0
(i.e. O uses terms that
have the namespace of O
0
). We provide an exhaustive
categorization of the adaptation to ontology evolution
for this situation. We observe these cases over two
ontology portals:
1. The Linked Open Vocabulary (LOV) ontology
portal which references 648 different ontologies,
88 of them evolved. We identified 74 cases of
ontology co-evolution, involving 28 different on-
tologies (Section 6.1).
2. The BioPortal which references 770 different on-
tologies, 485 of them evolved. We identified 14
cases of ontology co-evolution, involving 10 dif-
ferent ontologies (Section 6.2).
The usage of ontologies is increasing, so there is
the need of managing them, especially in the evolu-
tion process. The main aim of this research is to intro-
duce fundamentals for a methodological framework
for ontology management during the ontology evolu-
tion. Having such kind of frameworks will effectively
help to automate the process of managing ontologies
during their evolution cycle which can lead to save
time and effort.
We emphasize the need of having a service that
can automatically observe and notify the ontologies’
owners during the evolution process. Having such
tool can help to keep track of the different ontologies
during the co-evolution and help to facilitate the pro-
cess of ontology evolution.
REFERENCES
Abdel-Qader, M., Scherp, A., and Vagliano, I. (2018). An-
alyzing the evolution of vocabulary terms and their
impact on the LOD cloud. In The Semantic Web
- 15th International Conference, ESWC 2018, Her-
aklion, Crete, Greece, June 3-7, 2018, Proceedings,
pages 1–16.
Abgaz, Y. M., Javed, M., and Pahl, C. (2012). Analyz-
ing impacts of change operations in evolving ontolo-
gies. In Proceedings of the 2nd Joint Workshop on
Knowledge Evolution and Ontology Dynamics, USA,
November 12, 2012.
Brickley, D. and Miller, L. (2010). FOAF vocabulary spec-
ification 0.91.
Dragoni, M. and Ghidini, C. (2012). Evaluating the impact
of ontology evolution patterns on the effectiveness of
resources retrieval. In Proceedings of the 2nd Joint
Workshop on Knowledge Evolution and Ontology Dy-
namics, Boston, MA, USA, November 12, 2012.
Groß, A., Hartung, M., Pr
¨
ufer, K., Kelso, J., and Rahm, E.
(2012). Impact of ontology evolution on functional
analyses. Bioinformatics, 28(20):2671–2677.
Guha, R. V., Brickley, D., and Macbeth, S. (2016).
Schema.org: evolution of structured data on the web.
Communications of the ACM, 59(2):44–51.
Hartung, M., Groß, A., and Rahm, E. (2012). CODEX: ex-
ploration of semantic changes between ontology ver-
sions. Bioinformatics.
Kirsten, T., Hartung, M., Groß, A., and Rahm, E. (2009).
Efficient management of biomedical ontology ver-
sions. In On the Move to Meaningful Internet Sys-
tems: OTM 2009 Workshops, Confederated Interna-
tional Conferences, Portugal, pages 574–583.
Kupfer, A. and Eckstein, S. (2006). Coevolution of
database schemas and associated ontologies in biolog-
ical context. In 22nd British National Conference on
Databases.
Kupfer, A., Eckstein, S., Neumann, K., and Mathiak,
B. (2006). A coevolution approach for database
schemas and related ontologies. In 19th IEEE Interna-
tional Symposium on Computer-Based Medical Sys-
tems (CBMS 2006), Salt Lake City, Utah, USA, pages
605–610.
Lebo, T., Sahoo, S., and McGuinness, D. (2013). PROV-O:
The PROV ontology. W3C recommendation, 30.
Lehmann, J., Isele, R., Jakob, M., Jentzsch, A., Kon-
tokostas, D., Mendes, P. N., Hellmann, S., Morsey,
M., van Kleef, P., Auer, S., and Bizer, C. (2015). Db-
pedia - A large-scale, multilingual knowledge base ex-
tracted from wikipedia. Semantic Web, 6(2):167–195.
Mihindukulasooriya, N., Poveda-Villal
´
on, M., Garc
´
ıa-
Castro, R., and G
´
omez-P
´
erez, A. (2016). Collabora-
tive ontology evolution and data quality - an empirical
analysis. In 13th International Workshop, OWLED,
and 5th International Workshop, Bologna, Italy, pages
95–114.
Noy, N. F. and Musen, M. A. (2002). PROMPTDIFF:
A fixed-point algorithm for comparing ontology ver-
sions. In The Eighteenth National Conference on
Observing the Impact and Adaptation to the Evolution of an Imported Ontology
85

AI and Fourteenth Conference on Innovative Applica-
tions of AI, Canada.
Ottens, K., Aussenac-Gilles, N., Gleizes, M. P., and Camps,
V. (2007). Dynamic ontology co-evolution from texts:
Principles and case study. In Proceedings of the
First International Workshop on Emergent Semantics
and Ontology Evolution, ESOE 2007, co-located with
ISWC 2007 + ASWC 2007, Busan, Korea, pages 70–
83.
Papavassiliou, V., Flouris, G., Fundulaki, I., Kotzinos, D.,
and Christophides, V. (2009). On detecting high-level
changes in RDF/S kbs. In ISWC 2009, 8th Inter-
national Semantic Web Conference,USA., pages 473–
488.
Simperl, E. P. B. (2009). Reusing ontologies on the se-
mantic web: A feasibility study. Data Knowl. Eng.,
68(10):905–925.
Stojanovic, L., Maedche, A., Stojanovic, N., and Studer, R.
(2003). Ontology evolution as reconfiguration-design
problem solving. In Proceedings of the 2nd Inter-
national Conference on Knowledge Capture (K-CAP
2003), October 23-25, 2003, Sanibel Island, FL, USA.
Tartir, S., Arpinar, I. B., and Sheth, A. P. (2010). Onto-
logical evaluation and validation. In Theory and ap-
plications of ontology: Computer applications, pages
115–130. Springer.
Vandenbussche, P., Atemezing, G., Poveda-Villal
´
on, M.,
and Vatant, B. (2017). Linked open vocabularies
(LOV): A gateway to reusable semantic vocabularies
on the web. Semantic Web.
Whetzel, P. L., Noy, N. F., Shah, N., Alexander, P. R., Dorf,
M., Fergerson, R. W., Storey, M. D., Smith, B., Chute,
C. G., and Musen, M. A. (2011). Bioportal: Ontolo-
gies and integrated data resources at the click of a
mouse. In Proceedings of the 2nd International Conf.
on Biomedical Ontology, Buffalo, NY, USA.
Zablith, F., Antoniou, G., d’Aquin, M., Flouris, G., Kondy-
lakis, H., Motta, E., Plexousakis, D., and Sabou, M.
(2015). Ontology evolution: a process-centric survey.
Knowledge Eng. Review, 30(1):45–75.
KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development
86