=Paper=
{{Paper
|id=Vol-2477/paper_2
|storemode=property
|title=Updating Ontology Alignments in Life Sciences based on New Concepts and their Context
|pdfUrl=https://ceur-ws.org/Vol-2477/paper_2.pdf
|volume=Vol-2477
|authors=Victor Eiti Yamamoto,Julio Cesar dos Reis
|dblpUrl=https://dblp.org/rec/conf/semweb/YamamotoR19
}}
==Updating Ontology Alignments in Life Sciences based on New Concepts and their Context==
https://ceur-ws.org/Vol-2477/paper_2.pdf
Copyright ©2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
Updating Ontology Alignments in Life Sciences
based on New Concepts and their Context
Victor Eiti Yamamoto, Julio Cesar dos Reis
Institute of Computing, University of Campinas, Campinas - SP, Brazil
eitiyamamoto@gmail.com, jreis@ic.unicamp.br
Abstract. Ontologies and their associated mappings in life sciences play
a central role in several semantic-enabled tasks. However, the continuous
evolution of these ontologies requires updating existing concept align-
ments. Whereas mapping maintenance techniques have mostly handled
revision and removal type of ontology changes, the addition of concepts
demands further studies. This article proposes a technique to refine a set
of established mappings based on the evolution of biomedical ontologies.
We investigate ways of suggesting correspondences with the new version
of the ontology without applying a matching operation to the whole set
of ontology entities. Obtained results explore the neighbourhood of con-
cepts in the alignment process to update mapping sets. Our experimental
evaluation with several versions of aligned biomedical ontologies shows
the effectiveness in considering the context of new concepts.
Keywords: ontology alignment; ontology evolution; mapping refinement;
concept addition; biomedical vocabulary
1 Introduction
Over the last decade the biomedical domain has exploited ontologies and their
capabilities for various purposes ranging from information retrieval to data man-
agement and sharing. However, the size of this domain often requires the use of
several ontologies whose elements are linked through mappings. Mappings are
the materialization of semantic relations between elements of interrelated on-
tologies [14].
Creating mappings between ontologies is a complex task especially due to the
increasing size of biomedical ontologies. Several automatic ontology alignment
techniques have been proposed [10]. Nevertheless, significant manual efforts of
validation are still demanded if a certain level of quality is required. This prevents
software applications relying on mappings to fully take advantage on them.
Ontologies in life sciences evolve over time to keep them up-to-date accord-
ing to the domain knowledge. Ontology changes may affect mappings already
established or can be a source for treating mapping refinement. In this context,
in order to avoid the costly ontology re-alignment process, it is crucial to have
adequate mapping techniques to keep mappings semantically valid over time [1].
Manual mapping maintenance is possible only if modifications are applied to
16
2 Yamamoto and Dos Reis
a restricted number of mappings. Otherwise automatic methods are required
for large and highly dynamic ontologies. Biomedical ontologies usually contain
hundred of thousands of concepts interconnected via mappings.
Coping with the mapping reconciliation problem in a semi-automatic way
entails many research challenges. First, it is difficult to evaluate the real im-
pact of the ontology evolution on mappings. For instance, changing an attribute
value may lead to invalidate a mapping in some cases. In these situations, the
challenging issue is to identify and classify the different cases. Second, several
types of ontology changes can be applied to an ontology, but it is unknown how
these different types of operations should be duly taken into consideration for
mapping reconciliation [4].
The design of techniques for mapping adaptation according to the different
types of ontology changes has been coped within existing approaches. Previous
work presented a mapping adaptation strategy for two out of three categories
of ontology evolution: removal of knowledge and revision of knowledge [1]. For
example, when concepts are removed, heuristics were designed to automatically
apply adaptation actions over mappings. The addition of knowledge (third cat-
egory) is the most frequent type of change occurred in ontology evolution. New
concepts or attributes in concepts are added to comply with new domain knowl-
edge. Such new knowledge needs to be aligned with the interrelated ontologies.
In this paper, we propose a mapping refinement methodology to update map-
ping sets taking ontology changes into account (based on new concepts added
in ontology evolution). We study the use of conceptual information related to
neighbour concepts for enhancing the mapping completeness. For this purpose,
we investigate a technique to reuse already established mappings and to explore
the role of neighbour concepts to derive new mappings. Our proposal allows sug-
gesting new correspondences without applying a matching operation with the
whole set of ontology entities.
Our experimental evaluation explored real-world biomedical ontologies and
mappings established between them. We examine the quality of the automatically-
suggested enriched set of mappings with respect to the set of new correspon-
dences observed in the official updated release of mappings via standard evalu-
ation metrics. The achieved results show innovative findings regarding the way
mappings can be refined based on new concepts added. We demonstrate that the
local matching considering neighbour concepts is competitive with a matching
operation with the whole target ontology.
The remainder of this article is organized as follows: Section 2 presents the
related work; Section 3 presents the formal definitions and problem statement.
Section 4 reports on our approach to refine ontology mappings under ontology
evolution. Section 5 shows the used materials and the results obtained. Section
6 discusses the findings whereas Section 7 draws conclusions and future work.
17
Title Suppressed Due to Excessive Length 3
2 Related Work
Previous studies have investigated semi-automatic approaches to adapting ontol-
ogy mappings when at least one of the mapped ontologies evolves [4]. Dos Reis et
al. conceptualized the DyKOSMap framework [1] for supporting the adaptation
of semantic mappings highlighting different aspects such as: the role of different
types of ontology changes, the importance in considering the conceptual infor-
mation which established mappings are related to, as well as the relevance of the
different types of semantic relation of mappings.
Some techniques have used external resources aiming to improve and increase
the number and precision of established mappings. Stoutenburg [15] argued that
the use of upper ontologies (an ontology which consists of very general terms
that are common across all domains), and linguistic resources can enhance the
alignment process.
The TaxoMap matching tool [6] explored pattern-based refinement tech-
niques. The mapping is generated, with initial proposed relations (correspon-
dences found are equivalence relations, subsumption relations and their inverse,
or proximity relations). A domain expert manually validates the generated map-
pings and correct problems, grouping the identified problems together when they
correspond to a similar case. The tool generates patterns based on groups of sim-
ilar cases, which can be applied to other mappings in the same domain.
Other approaches have combined lexical-based and semantic-based algorithms,
mostly using resources available in the Unified Medical Language System (UMLS)1
for generating mappings. The use of UMLS as an external resource can be inter-
esting in various aspects: (1) favors an increase in the number of mapping, (2)
provides different synonyms terms for a given concept, and (3) defines relations
between concepts in a semantic network. Zhang and Bodenreider [16] explored
UMLS to improve alignment between anatomical ontologies. They showed that
domain knowledge is a key factor for the identification of additional mappings
compared with the generic schema matching approach.
Sekhavat and Parsons [13] explored conceptual models (e.g., Entity Relation-
ship, Class diagrams or domain ontologies) as background knowledge to enrich
database schema mappings and resolve ambiguous mappings. Their approach
used conceptual models as external resources to capture semantics of schema
elements, for instance, a pair of concepts a1 and a2 where a1 is a subclass and a2
is a superclass in a conceptual model. This information was used to enrich the
schema before mapping, marking the foreign keys corresponding to a1 and a2
as generalizations. As a consequence, the relationship identified in the schema
mapping is a generalization (is-a) instead of equivalence.
Pruski et al. [11] proposed exploiting domain-specific external source of knowl-
edge to characterize the evolution of concepts in dynamic ontologies. The tech-
nique analyzed the evolution of values in concept attributes. The approach used
1
UMLS is a collection of health and biomedical vocabularies and standards. URL:
www.nlm.nih.gov/research/umls/
18
4 Yamamoto and Dos Reis
ontological properties and mappings between ontologies from online repositories
to deduce the relationship between a concept and its successive version.
Noy et al. [9] and Seddiqui et al. [12] explored anchor concepts to obtain
mappings. They use a set of concept pairs aligned to obtain other mappings
based on these pairs. These approaches calculate new alignment for all concepts
from the involved ontologies, but they are not used for ontology evolution.
In this investigation, we explore ontology change operations to leverage re-
finement, in particular, concept addition. We contribute with a methodology to
consider newly added concepts and investigate the context of candidate target
concepts of existing mappings for refinement over time. We further evaluate the
proposed algorithms by measuring the effectiveness of our mapping refinement
approach on real-world biomedical ontologies.
3 Preliminaries
Ontology. An ontology O specifies a conceptualization of a domain in terms of
concepts, attributes and relationships [5]. Formally, an ontology O = (CO , RO , AO )
consists of a set of concepts CO interrelated by directed relationships RO . Each
concept c ∈ CO has a unique identifier and is associated with a set of attributes
AO (c) = {a1 , a2 , ..., ap }. Each relationship r(c1 , c2 ) ∈ RO is typically a triple
(c1 , c2 , t) where t is the relationship (e.g., “is a”, “part of”, “adviced by”, etc.)
interrelating c1 and c2 .
Context of a concept. We define the context of a particular concept ci ∈ CO
as a set of super concepts, sub concepts and sibling concepts of ci , as following:
CT (ci , λ) = sup(ci , λ) ∪ sub(ci , λ) ∪ sib(ci , λ) (1)
where
sup(ci , λ) ={cj |cj ∈ CO , r(ci , cj ) = “ @ ” ∧ length(ci , cj ) ≤ λ ∧ ci 6= cj }
sub(ci , λ) ={cj |cj ∈ CO , r(cj , ci ) = “ @ ” ∧ length(ci , cj ) ≤ λ ∧ ci 6= cj }
(2)
sib(ci , λ) ={cj |cj ∈ CO , ((sup(cj ) ∩ sup(ci )) ∨ (sub(cj ) ∩ sub(ci )))
∧length(ci , cj ) ≤ λ ∧ ci 6= cj }
where λ is the level of the context. It represents the maximum value for the
length between two concepts (in terms of their shortest relationship distance in
the hierarchy of concepts) and the “@” symbol indicates that “ci is a sub concept
of cj ”. This definition of CT (ci , λ) is specially designed as the relevant concepts
to be taken into account in the settings of this investigation on mapping refine-
ment.
Similarity between concepts. Given two particular concepts ci and cj , the
similarity between them can be defined as the maximum similarity between each
couple of attributes from ci and cj . Formally:
sim(ci , cj ) = arg max sim(aix , ajy ) (3)
19
Title Suppressed Due to Excessive Length 5
where sim(aix , ajy ) is the similarity between two attributes aix and ajy denoting
concepts ci and cj , respectively.
Mapping. Given two concepts cs and ct from two different ontologies, a map-
ping mst can be defined as:
mst = (cs , ct , semT ype, conf ) (4)
where semT ype is the semantic relation connecting cs and ct . In this article,
we differentiate relation from relationship, where the former belongs to a map-
ping and the later to an ontology. The following types of semantic relation
are considered: unmappable [⊥], equivalent [≡], narrow-to-broad [≤], broad-to-
narrow [≥] and overlapped [≈]. For example, concepts can be equivalent (e.g.,
“head”≡“head”), one concept can be less or more general than the other (e.g.,
“thumb”≤“finger”) or concepts can be somehow semantically related (≈). The
conf is the similarity between cs and ct indicating the confidence of their rela-
tion [3]. We define MjST as a set of mappings mst between ontologies OS and
OT at a given time j. We assume j ∈ N the version of the ontology release OSj .
Ontology OS0 is the version 0 whereas OS1 is the version 1 of the same ontology.
Ontology change operations (OCO). An ontology change operation (OCO)
is defined to represent a change in an attribute, in a set of one or more concepts
or in a relationship between concepts. OCOs are classified into two main cate-
gories: atomic and complex changes. Each OCO in the former cannot be divided
into smaller operations while each one of the latter is composed of more than
one atomic operation. In this paper, we pay further attention to the operations
of concept addition which is an atomic operation.
Problem statement. Consider two versions of the same source ontology
OSj at time j and OSj+1 at time j + 1, a target ontology OTj , and an initial set of
mappings MjST between OSj and OTj at time j. Suppose that the frequency of new
releases of OS and OT is different and at time j + 1 only OS evolves. Since this
evolution is likely to impact the mappings MjST , it is necessary to refine MjST
to guarantee the quality and completeness of Mj+1ST . The quality is related to the
consistency of mappings and it can be measured using precision. For instance,
mappings cannot be established between removed concepts. The completeness
refers to the recall of aligned concepts in Mj+1
ST . In this investigation, we study
how MjST can be refined (e.g., new mappings derived) based on ontology changes
related to addition of knowledge. We address the following research questions:
• How to exploit existing mappings for mapping refinement based on new
concepts added?
• Is it possible to reach mapping refinement for alignment of new concepts
without applying a matching operation in the whole target ontology?
• What is the impact of using the context of concepts CT (ci , λ) in both source
and target ontologies on the mapping refinement effectiveness?
We consider that OT has not evolved (thus OTj and OTj+1 are the same version
of the ontology OT ). OSj and OSj+1 are two distinct versions of the same ontology
20
6 Yamamoto and Dos Reis
OS . At time j + 1, newly added concepts appear in OSj+1 and we attempt to
j j+1
refine the original mapping set MST to provide a set of valid mappings MST .
4 Mapping Refinement under Concept Addition Changes
Our goal is to propose adequate correspondences for each newly added concept
at time j + 1. In the first step, our approach identifies all newly added concepts
using the Conto-Diff tool [7]. This tool allows to identify atomic and complex
ontology changes. Next, we extract the contextual information, i.e., super, sub
and sibling concepts of those newly added concepts (cf. Formula 1). We then
examine the existing mappings between the source concept in the context of the
newly added concept and the corresponding target concepts. The idea behind the
context-oriented technique is that the candidate mapping is established between
a newly added concept and a target concept of an existing mapping at time t.
Figure 2 A illustrates a situation where there are two ontologies that have an
alignment in time j. Each circle represents a concept of an ontology. Light blue
circles represent concepts of source ontology. Yellow circles represent concepts
of target ontology. Continuous lines represent mappings between concepts from
source ontology and target ontology.
Figure 2 B illustrates a situation where source ontology evolve and change
to time j + 1. The algorithm find newly added concepts and explore the context
of each newly added concepts. In this case, we are exploring the context of the
right concept using source level 1. Purple circles represent newly added concepts.
Dark blue circles represent concepts of a context with a certain source level; the
number inside the circle represents the source level needed to access this concept.
After finding some concepts inside the newly added concepts’ context that
have an alignment in previous time, the concepts from target ontologies that have
an alignment in previous time are added as candidate concepts. The context of
each candidate concepts is explored and added as candidate concepts. Figure
3 illustrate this situation using target level 1. Red circles represent candidate
concepts for a new concept from source ontology; the number inside the circles
represent the target level needed to access this concept. Dashed lines represent
a possible alignment between a new concept (in OSj+1 ) and candidate concepts
(in OTj ).
Algorithm 1 computes the diff between two given versions of the source
ontology (line 1). For each newly added concept c1i , the algorithm considers a
candidate concept namely c0t in the target ontology by exploiting already exist-
ing mappings related to CT (c1i , γ) (lines 4-8). Note that we recover the before
evolution version (c0k ) of the concept c1k found in the context of c1i .
For each c0t , the algorithm obtains a set of concepts from CT (c0t , λ) (line 11).
We determine a new refined mapping by calculating the similarity between a new
concept c1i of OSj+1 and a candidate cn ∈ Ct . If the maximum similarity (among
the concept attributes) is greater than or equal to a threshold τ , the algorithm
establishes a mapping between the newly added concept and the candidate target
21
Title Suppressed Due to Excessive Length 7
A. Initial situation B. Finding source context
Fig. 2: Representation of situations before applying alignment algorithm
Fig. 3: Calculating similarity with candidate concepts
concept. Algorithm 1 searches for the candidate ct that yields the maximum
similarity value.
In order to compare with the results obtained by our approach (cf. Section5),
we propose another algorithm that ignores new concepts’ context to calculate
similarity. It means that the algorithm computes the similarity between each
newly added concepts with all concepts in the target ontology. More specifically,
these algorithm computes the diff between two given versions of the source ontol-
ogy. For each newly added concepts, it calculates similarity between all concepts
with the target ontology. If there are any similarity greater than a threshold,
22
8 Yamamoto and Dos Reis
Algorithm 1 Contextual approach to mapping refinement
Require: OSj , OSj+1 , OTj , OTj+1 , MjST , λ, γ, τ ∈ R
Ensure: MA = {m1 , m2 , ..., mN }
1: Cadd ← dif fadd (OSj , OSj+1 ) {newly added concepts}
2: Ct ← ∅ {initialize target concepts of candidate mappings}
3: for all c1i ∈ Cadd do
4: for all c1k ∈ CT (c1i , γ) do
5: if ∃c0t ∈ CO0 , ∃m(c0k , c0t ) ∈ M0ST then
T
{c0t }
S
6: Ct ← Ct
7: end if
8: end for
9: mit ← ∅
10: for all ct ∈ Ct do
11: for all cn ∈ CT (ct , λ) do
12: mcand ← argmax sim(c1i , cn ) {Create a mapping between concepts c1i and
cn }
13: if max(sim(c1i , cn )) ≥ τ then
14: mit ← mcand
15: τ ← max(sim(c1i , cn ))
16: end if
17: end for
18: end for S
19: MA ← MA {mit }
20: end for
21: return MA
the algorithm creates a new mapping between the newly added concept and a
target ontology’s concept with the greatest similarity.
In our algorithms, source concept’s attributes are compared with all target
concept’s attributes to obtain similarity value between concepts. The value of
similarity between two concepts is the maximum value of similarity from their
attributes. The method used to calculate similarity affects the precision and
recall. In this work, we explored Bi-gram Dice to calculate similarity. Bi-gram is
a sequence of two adjacent letters of a word. Dice’s coefficient is defined as twice
the number of common elements divided by sum of each elements. Formula 5
shows the application of Bi-gram Dice to strings X and Y. The n-gram’s strength
is in the fact it has context sensitivity, but it not have good resolution when gram
size is increased [8]. For the data set used in this work Bi-gram Dice have better
results than Levenshtein distance, Cosine distance and Jaccard distance.
2 × (Bi − gram(X) ∩ Bi − gram(Y )
Similarity = (5)
Bi − gram(X) + Bi − gram(Y )
23
Title Suppressed Due to Excessive Length 9
5 Experimental Evaluation
We aim to validate the quality of the refined set of mappings as the outcome
of our approach. Data used in this evaluation come from five biomedical on-
tologies: SNOMED-CT (SCT), MeSH, ICD-9-CM, ICD10-CM and NCI The-
saurus. SNOMED-CT (Systematized Nomenclature of MedicineClinical Terms)
is an ontology which objective is to create a taxonomy of terms referring to
the medical environment and a framework of rules guaranteeing that each term
is used with exactly one meaning [2]. MeSH Thesaurus is a controlled vocab-
ulary produced by the National Library of Medicine and used to index, cata-
logue and search information and documents related to biomedicine and health
https://www.nlm.nih.gov/ ICD-9-CM and ICD-10-CM are a formalization in
OWL-DL of International Classification of Diseases published by World Health
Organization2 NCI Thesaurus3 contains terminologies used in the National Can-
cer Institute’s semantic infrastructure and information systems Table 1 shows
the statistics of source and target ontologies for each of the considered versions.
Table 1: Statistics of ontologies
Ontology Release #Concepts #Attributes #Subsumptions #New Concepts
ICD9 2009 12,734 34,065 11,619 325
2011 13,059 34,963 11,962
ICD10 2011 43,351 87,354 40,330 0
SCT 2010 386,965 1,531.288 523,958 8,381
2012 395,346 1,570,504 539,245
NCI 2009 77,448 282,434 86,822 17,284
2012 94,732 365,515 105,406
MeSH 2012 50,367 259,565 59,191 604
2013 50,971 264,783 59,844
The mappings obtained by the proposed Algorithm 1 are compared with the
official mappings (their new official release). Mappings between SNOMEDCT
and ICD9CM is offered by the International Health Terminology Standards De-
velopment Organisation (IHTSDO)4 . Mappings between MeSH and ICD-10-CM
were offered by the Catalogue et Indexation des Sites Mdicaux de langue Franaise
(CISMeF)5 . Table 2 shows the quantity of each mapping set between the ontolo-
gies used in this experiment.
To analyze results obtained experimentally, it was necessary to compare our
obtained mappings with mappings created only for newly added concepts in the
new version of the considered ontologies. Table 3 shows the quantity of mappings
really considered in the metrics.
2
http://www.who.int/classifications/icd/en/
3
https://ncit.nci.nih.gov/ncitbrowser/
4
ttps://www.nlm.nih.gov/research/umls/mapping projects/icd9cm to snomedct.html
5
http://www.chu- rouen.fr/cismef
24
10 Yamamoto and Dos Reis
Table 2: Statistics of the studied mappings
SCT-ICD9 #Mappings SCT-NCI #Mappings MeSH-ICD10CM #Mappings
2010-2009 84,519 2009-2009 19,971 2012-2011 4,631
2012-2011 86,638 2012-2012 22,732 2013-2011 5,378
Table 3: Number of new mappings created and associated to newly added concepts in
the new ontology version (considered gold standard)
Mappings #official mappings created after newly added concepts
SNOMEDCT-ICD9CM 1,583
SNOMEDCT-NCI 158
MeSH-ICD10CM 21
The experiments were performed for the three datasets (SCT-NCI, SCT-
ICD9 and MeSH-ICD10) considering SCT and MeSH as source ontologies. As
assessed configurations, we considered three source levels, three threshold values
(0.5, 0.75 and 0.9), and four target levels. For each dataset, we fixed source
level and threshold to verify the results for each target level. After examining all
target levels, we changed the threshold and repeated for each target level. After
examining all threshold values, we changed source level and repeated the whole
procedure for all thresholds and target levels.
We used three metrics to evaluate the results: Precision, Recall and F-
Measure. These metrics were used comparing results obtained by our approach
and expected results from the official mappings.
Precision is defined as the relation between correctly identified mappings and
identified mapping (Formula 6).
#Identif edAndCorrectM appings
P recision = (6)
#Identif iedM appings
Recall is defined as the relation between correctly identified mappings and
those expected new official release of mappings (Formula 7).
#Identif edAndCorrectM appings
Recall = (7)
#CorrectM appings
F-measure is the harmonic mean of precision and recall (Formula 8).
2 × P recision × Recall
F − M easure = (8)
P recision + Recall
Tables 4 (SNOMED-CT and NCI Thesaurus), 5 (SNOMED-CT and ICD-9)
and 6 (MeSH and ICD-10) show the obtained results in terms of precision, recall
and f-measure in applying our Algorithm 1 for the studied datasets.
Results in Table 4 reveal a decrease in precision and f-Measure for threshold
set as 0.5 when it increases source level. The results increases in terms of preci-
sion, recall and F-measure for other thresholds. We found that the best results
are obtained when increasing the level of the context in the source concept.
25
Title Suppressed Due to Excessive Length 11
Table 4: Mapping derivation results for SNOMED-CT and NCI
Source level Threshold Target level Precision Recall F-Measure
1 0.5 0 0.009 0.018 0.012
1 0.042 0.101 0.060
2 0.048 0.120 0.068
3 0.048 0.127 0.069
0.75 0 0 0 0
1 0.088 0.082 0.085
2 0.086 0.082 0.0841
3 0.101 0.108 0.104
0.9 0 0 0 0
1 0.344 0.070 0.116
2 0.378 0.089 0.144
3 0.341 0.089 0.140
2 0.5 0 0.006 0.025 0.010
1 0.031 0.139 0.051
2 0.035 0.171 0.058
3 0.034 0.184 0.058
0.75 0 0.006 0.006 0.006
1 0.089 0.120 0.102
2 0.105 0.152 0.124
3 0.108 0.165 0.130
0.9 0 0.071 0.006 0.012
1 0.357 0.095 0.012
2 0.423 0.127 0.195
3 0.407 0.139 0.208
3 0.5 0 0.004 0.019 0.006
1 0.023 0.139 0.039
2 0.029 0.190 0.050
3 0.028 0.190 0.048
0.75 0 0.005 0.006 0.006
1 0.08 0.127 0.097
2 0.104 0.177 0.125
3 0.097 0.177 0.125
0.9 0 0.071 0.006 0.012
1 0.356 0.101 0.158
2 0.434 0.146 0.218
3 0.418 0.146 0.216
Results presented in Table 5 (concerning the mapping between SNOMED-
CT and ICD-9) are different from those of SNOMED-CT and NCI. We observe
an increase in recall when it increases source level, but it has lower precision.
Table 5 presents the best results for the first level in the source concept and
with lower thresholds. We could not observe huge differences in the results when
increasing the context level of the target concept.
Table 6 presents the results for the refinement for MeSH and ICD-10. We
observed an overall improvement of results when increasing the level of the source
concept.
We evaluated our proposal in considering the neighbourhood for the deriva-
tion of new mappings associated to new concepts (Algorithm 1) with the ap-
proach in applying the matching with the whole target ontology. To this end,
we applied the non-context approach in the datasets considering the threshold τ
yielding the best results in Algorithm 1 obtained for each dataset. Table 7 shows
the results concerning precision, recall and f-measure obtained for each dataset
using the matching with all concepts in the target ontology. The comparison of
26
12 Yamamoto and Dos Reis
Table 5: Mapping derivation results for SNOMED-CT and ICD-9
Source level Threshold Target level Precision Recall F-Measure
1 0.5 0 0.535 0.186 0.276
1 0.340 0.163 0.220
2 0.310 0.152 0.204
3 0.296 0.145 0.196
0.75 0 0.630 0.037 0.069
1 0.461 0.044 0.081
2 0.449 0.041 0.075
3 0.439 0.041 0.075
0.9 0 0.778 0.004 0.009
1 0.692 0.006 0.011
2 0.727 0.005 0.10
3 0.75 0.006 0.011
2 0.5 0 0.325 0.181 0.233
1 0.233 0.159 0.189
2 0.241 0.167 0.198
3 0.230 0.160 0.230
0.75 0 0.487 0.046 0.084
1 0.349 0.0455 0.080
2 0.449 0.042 0.076
3 0.382 0.048 0.085
0.9 0 0.8 0.005 0.010
1 0.687 0.007 0.014
2 0.615 0.005 0.010
3 0.615 0.005 0.010
3 0.5 0 0.256 0.177 0.209
1 0.190 0.158 0.166
2 0.200 0.159 0.177
3 0.199 0.157 0.175
0.75 0 0.444 0.051 0.091
1 0.342 0.049 0.085
2 0.360 0.050 0.089
3 0.348 0.049 0.085
0.9 0 0.833 0.006 0.013
1 0.687 0.007 0.014
2 0.687 0.007 0.014
3 0.687 0.007 0.014
results reveals that for the dataset SCT-NCI the results using all target concept
as candidates were better. For the dataset SCT-ICD9, our context-approach is
better; concerning the dataset MeSH-ICD10, the approaches obtained similar
results. However, we need to consider that applying mapping candidates with
the whole target ontology have a worst run-time complexity than our contextual
approach.
6 Discussion
This investigation aimed to create mappings to update ontology alignments
based on new concepts added in novel ontology releases. Our approach have
three variables affecting mapping quality: threshold, target level and source level.
Threshold increases precision, but decreases recall. It is caused by the fact that
high threshold can remove false positive mappings, but as an effect removes cor-
rect mappings. For two datasets (SCT-NCI and MeSH-ICD10) the increasing in
27
Title Suppressed Due to Excessive Length 13
Table 6: Mapping refinement results for MeSH and ICD-10
Source level Threshold Target level Precision Recall F-Measure
1 0.5 0 0.059 0.048 0.053
1 0.059 0.048 0.053
2 0.059 0.048 0.053
3 0.0625 0.048 0.054
0.75 0 0.250 0.048 0.08
1 0.250 0.048 0.08
2 0.250 0.048 0.08
3 0.250 0.048 0.08
0.9 0 0 0 0
1 0 0 0
2 0 0 0
3 0 0 0
2 0.5 0 0.067 0.095 0.078
1 0.079 0.143 0.102
2 0.0714 0.143 0.095
3 0.0714 0.143 0.095
0.75 0 0.250 0.048 0.08
1 0.429 0.143 0.214
2 0.429 0.143 0.214
3 0.429 0.143 0.214
0.9 0 0 0 0
1 1.000 0.095 0.174
2 1.000 0.095 0.174
3 1.000 0.095 0.174
3 0.5 0 0.036 0.095 0.052
1 0.044 0.143 0.067
2 0.043 0.143 0.067
3 0.043 0.143 0.066
0.75 0 0.125 0.048 0.069
1 0.333 0.143 0.2
2 0.333 0.143 0.2
3 0.333 0.143 0.2
0.9 0 0 0 0
1 1.000 0.095 0.174
2 1.000 0.095 0.174
3 1.000 0.095 0.174
Table 7: Mapping derivation results exploring the matching with all concepts of the
target ontology
Data set Threshold Precision Recall F-Measure
SCT-NCI 0.9 0.593 0.525 0.557
SCT-ICD9 0.5 0.264 0.042 0.072
MeSH-ICD10 0.75 0.312 0.238 0.270
precision compensated decreasing in recall. However, we observed for one dataset
(SCT-ICD9) that a high threshold implied bad effects.
Target level increases candidate concepts for mapping by increasing the con-
text in the target ontology. It means that each new added concept has more
options to compare. The increasing in candidate concepts means more chances
to find a correct mappings, but it can cause finding a wrong mapping when a
wrong concept have a better results in terms of similarity value than the expected
concept. For two datasets (SCT-NCI and MeSH-ICD10), precision increased be-
tween target level 0 and 1 and recall improved when target level increases. For
28
14 Yamamoto and Dos Reis
one dataset (SCT-ICD9), precision decreased between target level 0 and 1 and
recall had only minor effects caused by changes in target level. We found that
the results were very dependent on the characteristics of the datasets.
Source level increases source context to find candidate concepts from the
target ontology. Our approach depends if in the neighbourhood of a new concept
there are concepts presenting a mapping in prior version. If the source level is low,
new concepts have less chances to find concepts mapped in prior version. In the
worst case, if there is no concept mapped in a prior version, the new concepts are
not analyzed to find new mappings. Therefore, in our approach, the derivation
of mappings related to new concepts depends directly on the source level. We
found better results in improving the level of the source concept.
The analysis of results obtained by all concepts approach indicates that SCT-
NCI got better results using such approach. Whereas precision presents very
similar results, the recall is very low using contextual approach. SCT-ICD9 pre-
sented better results using contextual approach. In this case, precision had good
values using contextual approach, but f-measure suffered with low recall. Find-
ings on the dataset MeSH-ICD10 presented similar results for both approaches.
In summary, contextual approach implies in a better precision, but all concepts
approach obtains a better recall.
This research found that is possible to exploit existing mappings for mapping
refinement based on new concepts added. Our findings indicated the possibility of
reaching mapping refinement for alignment of new concepts in ontology evolution
without applying a matching operation in the whole target ontology. We found
further impact in considering the level of the source concept than in the target
ontology for the effectiveness of the ontology alignment refinement.
7 Conclusion
Ontology mappings play a central role for semantic data integration in the life
sciences. However, domain knowledge update leads to new concepts in ontology
versions. This requires to maintain mapping sets up-to-date according to the
knowledge dynamics. We proposed a technique to refine ontology alignments
based on evolving ontologies. Our constructed algorithm considered the context
of concepts in both ontologies as a way to find the matching between concepts.
Experimental evaluation with aligned ontologies in the life sciences demonstrated
the effectiveness of our approach. Future work involves further investigating
heuristics to update the type of semantic relation in the refinement procedure.
Acknowledgements
This work was financially supported by the São Paulo Research Foundation
(FAPESP) (grant #2017/02325-5)6 . We also thank the Scientific Initiation Pro-
gram (PIBIC) from UNICAMP for the scholarship grant.
6
The opinions expressed in here are not necessarily shared by the financial support
agency.
29
Title Suppressed Due to Excessive Length 15
References
1. Dos Reis, J.C., Pruski, C., Da Silveira, M., Reynaud-Delaitre, C.: Dykosmap: A
framework for mapping adaptation between biomedical knowledge organization
systems. Journal of Biomedical Informatics 55, 153 – 173 (2015)
2. El-Sappagh, S., Franda, F., Ali, F., Kwak, K.: Snomed ct standard ontology based
on the ontology for general medical science. BMC Medical Informatics and Decision
Making 18 (12 2018)
3. Euzenat, J., Shvaiko, P.: Ontology matching. Springer (2007)
4. Groß, A., Reis, J.C.D., Hartung, M., Pruski, C., Rahm, E.: Semi-Automatic Adap-
tation of Mappings between Life Science Ontologies. In: Proceedings The 9th In-
ternational Conference on Data Integration in the Life Sciences. pp. 90–104 (2013)
5. Gruber, T.R.: A translation approach to portable ontology specifications. Knowl.
Acquis. 5(2), 199–220 (Jun 1993)
6. Hamdi, F., Safar, B., Niraula, N.B., Reynaud, C.: Taxomap alignment and re-
finement modules: Results for oaei 2010. In: Proceedings of the 5th International
Workshop on Ontology Matching. vol. 689, pp. 212–219 (2010)
7. Hartung, M., Gross, A., Rahm, E.: COnto-Diff: Generation of Complex Evolution
Mappings for Life Science Ontologies. Biomedical Informatics 46, 15–32 (2013)
8. Kondrak, G.: N-gram similarity and distance. In: Consens, M., Navarro, G. (eds.)
String Processing and Information Retrieval. pp. 115–126. Springer Berlin Heidel-
berg, Berlin, Heidelberg (2005)
9. Noy, N.F., Musen, M.A.: Anchor-prompt: Using non-local context for semantic
matching. In: Workshop on ontologies and information sharing. pp. 63–70 (2001)
10. Otero-Cerdeira, L., Rodrguez-Martnez, F.J., Gmez-Rodrguez, A.: Ontology match-
ing: A literature review. Expert Systems with Applications 42(2), 949 – 971 (2015)
11. Pruski, C., Dos Reis, J.C., Da Silveira, M.: Capturing the relationship between
evolving biomedical concepts via background knowledge. In: Proceedings of the
9th International Conference on Semantic Web Applications and Tools for Life
Sciences (SWAT4LS16) (2016)
12. Seddiqui, M.H., Aono, M.: Anchor-flood: Results for oaei 2009. In: Proceedings of
the 4th International Conference on Ontology Matching - Volume 551. pp. 127–134.
OM’09, CEUR-WS.org, Aachen, Germany, Germany (2009)
13. Sekhavat, Y.A., Parsons, J.: Sesm: Semantic enrichment of schema mappings. In:
Proceedings of the 29th International Conference on Data Engineering Workshops
(ICDEW 2013). pp. 7–12. IEEE (2013)
14. Shvaiko, P., Euzenat, J.: Ontology Matching: State of the Art and Future Chal-
lenges. IEEE Trans. Knowl. Data Eng. 25(1), 158–176 (2013)
15. Stoutenburg, S.K.: Acquiring advanced properties in ontology mapping. In: Pro-
ceedings of the 2nd PhD Workshop on Information and Knowledge Management
(PIKM 2008). pp. 9–16. ACM (2008)
16. Zhang, S., Bodenreider, O.: Experience in aligning anatomical ontologies. Interna-
tional journal on Semantic Web and information systems 3(2), 1 (2007)
30