A USER-DRIVEN AND A SEMANTIC-BASED ONTOLOGY
MAPPING EVOLUTION APPROACH
Hélio Martins and Nuno Silva
GECAD - Knowledge Engineering and Decision Support Research Group
School of Engineering – Polytechnic of Porto, Porto, Portugal
Keywords: Ontology, Ontology Evolution, Ontology Mapping, Ontology Mapping Evolution.
Abstract: Systems or software agents do not always agree on the information being shared, justifying the use of
distinct ontologies for the same domain. For achieving interoperability, declarative mappings are used as a
basis for exchanging information between systems. However, in dynamic environments like the Web and
the Semantic Web, ontologies constantly evolve, potentially leading to invalid ontology mappings. This
paper presents two approaches for managing ontology mapping evolution: a user-centric approach in which
the user defines the mapping evolution strategies to be applied automatically by the system, and a semantic-
based approach, in which the ontology’s evolution logs are exploited to capture the semantics of changes
and then adapted to (and applied on at) the ontology mapping evolution process.
1 INTRODUCTION
Ontologies are a key concept in knowledge-based
systems in general and in the Semantic Web in
particular (Berners-Lee et al., 2001). However,
actors do not always agree on the information being
shared, justifying the use of distinct ontologies, even
if corresponding to the same domain of application.
Information integration arises therefore as a core
process in these application areas. To
solve/minimize the interoperability problem,
ontology mapping proved to be an efficient solution
(Silva, 2004). Ontology mapping is the process
whereby semantic relations are defined between two
ontologies at conceptual level, which in turn are
applied at data level, transforming source ontology
instances into target ontology instances (Silva,
2004). The result of the ontology mapping
specification process (at conceptual level) is an
ontology mapping document containing the semantic
relations. However, in dynamic environments,
ontologies evolve, causing the ontology mapping
document to become invalid. In heterogeneous
environments where interoperability among systems
depends essentially upon the ontology mappings,
semantic relations must be adapted to reflect the
ontologies evolution (Figure 1). The two original
ontologies (O1 and O2) are mapped through the M
mapping document (more than two ontologies may
exist in the integration scenario, but for the sake of
simplicity, only two are considered here).
Figure 1: Problem definition scenario.
However, O1 and/or O2 may evolve, giving rise
to O1’ and O2’ respectively. This process is named
ontology evolution, and may cause inconsistencies
in the ontology mapping document. To solve those
inconsistencies, two solutions were identified: (i)
generation of a completely new ontology mapping
document, i.e. a new set of semantic relations
between entities of versions of source/target
ontologies and (ii) adapt the original ontology
mapping document to the ontologies’ evolution, i.e.
correcting invalid semantic relation and generating
214
Martins H. and Silva N. (2009).
A USER-DRIVEN AND A SEMANTIC-BASED ONTOLOGY MAPPING EVOLUTION APPROACH.
In Proceedings of the 11th International Conference on Enterprise Information Systems - Databases and Information Systems Integration, pages
214-221
DOI: 10.5220/0002011002140221
Copyright
c
SciTePress
new required semantic relations according to
ontology changes, giving rise to a new ontology
mapping document (M’). Considering all the
difficulties and problems faced by the ontology
mapping process (Euzenat, et al., 2007), it is worth
considering the second approach.
The goal of this work is therefore to research and
develop a solution for supporting the evolution and
adaptation of the ontology mapping document
according to the mapped ontologies’ evolution.
In the reminder of the paper, we present the
context of this work in Section 0. In Section 0 we
present the foundations of our ontology mapping
evolution process. In Section 0 a user-driven
approach is presented, followed by the description of
the semantic-based approach in Section 0. In Section
0 we describe the evaluation experiences. In Section
0, the related work and provide our concluding
remarks in Section 0.
2 CONTEXT
2.1 Ontology Evolution
Ontology evolution can be defined as the timely
adaptation of ontologies to changes which arises and
the consistent management of these changes
(Stojanovic, 2004). According to this definition, two
key concepts were indentified: (i) ontology changes
and (ii) consistency management.
Different levels of change abstraction are often
used (Stojanovic, 2004), (Klein, 2004), (Plessers,
2005), which might be summarized into:
Basic changes, that change one entity in the
ontology model only (e.g. Removing a Concept from
the ontology);
Composite changes, that change more than one
entity in the ontology (e.g. Copy a Concept).
These changes are part of an evolution ontology
that conceptualizes all kind of changes that can be
performed on the ontology, the relation between
them and meta-information (e.g. date, owner of
change request). During the ontology evolution
process, the changes are stored in an instance of the
evolution ontology, serving as ontology evolution
log, e.g.:
<RemoveEntity
rdf:ID="i-151"
referencesConcept="http://s1.pt#C2">
<causesChange rdf:resource="#i-66"/>
<causesChange rdf:resource="#i-92"/>
<previousChange rdf:resource="#i-99"/>
</RemoveEntity>
Consistency defines the degree of uniformity,
standardization, and freedom from contradiction
among the parts of a system or component (IEEE,
1990). In (Stojanovic, 2004) the author states that
ontology consistency can be considered as an
agreement among ontology entities with the respect
to the semantics of the underlying ontology
language. This assumes special relevance because
the execution of a single ontology change (e.g.
removing a concept) may cause inconsistencies in
other parts of the ontology (e.g. subClassOf relation
between concepts). To solve these inconsistencies,
many approaches use the concept of derived or
deduced changes. Because deduced changes may
result in additional deduced changes, this becomes
an iterative process. As consequence, it is relevant to
distinguish between the changes requested by the
ontology engineer (representing their intentions) and
the deduced changes (to solve inconsistencies).
Notice that beyond the agreement of the semantics
of the ontology language, ontology consistency is
related with the ontology’s purpose. Yet, this
dimension is not addressed in this paper.
Complementarily, the concept of ontology
evolution strategy proposed in (Stojanovic, 2004)
assumes special relevance in the context of this
work. The evolution strategy concept is used to give
the user (i.e. the ontology engineer) the possibility to
achieve a consistent ontology according to the
semantics of the ontology language and a set of best-
practices. This concept is based on:
One resolution point, i.e. a dilemma that might
occur during the resolution of changes;
Elementary ontology evolution strategies, i.e.
the potential ways for resolving one resolution point.
A pair between one resolution point and one
elementary strategy is named ontology evolution
strategy. Table 1 presents the resolution points and
respective elementary evolution strategies that are
useful for this work. For example, if a concept
becomes orphan (i.e. no subClassOf relation remains
for the concept), one may choose to: (i) delete the
concept, reconnect the concept to the parent of the
removed concept, or reconnect it to the root concept.
Table 1: Ontology evolution strategies.
Evolution Strategy
Resolution Point Elementary strategy
Orphaned concept
Delete
Reconnect to parent
Reconnect to root
Property propagation
Don’t Propagate
Propagate direct only
Propagate all
A USER-DRIVEN AND A SEMANTIC-BASED ONTOLOGY MAPPING EVOLUTION APPROACH
215
Equivalent reasoning is made for the other
resolution point. Evolution strategies will be
discussed in section 0.
2.2 Ontology Mapping
Ontology mapping proved to be an efficient solution
(Rahm, et al., 2001), (Euzenat, et al., 2007) for
helping solve the interoperability problem. The
IEEE dictionary (IEEE, 1990) defines
interoperability as “the ability of two or more
systems or components to exchange information and
to use the information that has been exchanged”.
The ontology mapping process is not a trivial
process, requiring a deep understanding of both
ontology conceptualizations and their semantic
similarities. The output of the ontology mapping
process is an ontology mapping document
containing semantic relations between source
ontology entities and target ontology entities. In our
context, an ontology mapping document is an
instantiation of the SBO (Semantic Bridge
Ontology) (Silva, 2004). The SBO is the ontology
that describes the semantic relations holding
between ontologies entities, providing not only a
conceptualization mechanism but also a
representation and exchange mechanism of semantic
relationships between ontologies. It specifies,
classifies and describes the types of ontology
mapping relations, inter relates them and provides
other modelling constructs necessary to express
ontology mapping documents.
The SBO is fundamentally composed of the
following entities:
Concept Bridge is the semantic relation that
maps ontologies’ concepts. At transformation time,
it will create target concept instances. For example:
CB(O1:Person, O2:Individual), will create an
instance of O2:Individual for each existing
O1:Person;
Property Bridge, maps source properties to
target properties. At execution time, properties are
created in the scope of the target instance. For
example, PB({O1:firstName,
O1:lastName},{O2:name},concat) will concatenate
the value of O1:firstName and O1:lastName into the
value of O2:name. The service (e.g. Concat,
CopyAttribute) defines the arguments of the
Property Bridge;
hasBridge relation associates a PropertyBridge
to a ConceptBridge. This relation provides the scope
to the Property Bridge. Otherwise, the Property
Bridge would not realize to which concept the
property value is to be attached;
subBridgeOf relation between ConceptBridges
is the mapping equivalent to the ontological
“subClassOf” relation.
The ontology mapping process is responsible for
the instantiation of the SBO, generating a mapping
document. Figure 2 illustrates an example of a
mapping document.
Figure 2: Example of ontology mapping document.
Once instantiated, the semantic bridges can be
exploited at data level (instance level) for
information transformation/exchange.
3 MAPPING EVOLUTION
Ontology mapping evolution is the process whereby
entities of ontology mapping document are adapted
with the eventual changes in the source and/or target
ontologies, trying to preserve as much of the
semantics of the original mapping relations as
possible.
3.1 Two-phases Iterative Process
Changes to ontologies affect the ontology mapping
in two ways: whether it is an addition or a removal.
In the case of removal, existing semantic
relationships are affected and must evolve. Instead,
if adding new ontology entities, new semantic
relations might be necessary.
Accordingly, the proposed ontology mapping
evolution process comprises of two iterative phases:
Correcting Invalid Entities, consists of
identifying and correcting the invalid mapping
entities according to the ontologies evolution,
preserving as much as possible the semantics of the
original mapping;
ICEIS 2009 - International Conference on Enterprise Information Systems
216
Matching New Ontology Entities, consists of
discovering and evaluating the similarity between
entities and the specification of (correct) mapping
relationships.
Matching new entities is closely related to the
matching process (Euzenat, et al., 2007) and less (or
nothing) with the mapping evolution process, thus
will not be addressed further in this paper. Instead,
the following sections concern the process of
Correcting Invalid Entities.
3.2 The Semantics of SBO
In order to correct an invalid SBO entity, one has to
identify it first. The entity is invalid when the SBO
semantics are not obeyed. Depending on the SBO
entity, different semantic restrictions apply.
An invalid Concept Bridge is one that has at
least one invalid argument (Table 2).
Table 2: Concept Bridge’s invalid conditions.
Invalid Argument Causes
Source Concept
It is not a concept, or
The concept value does not exist, or
It is not a source ontology’s concept
Target Concept
It is not a concept, or
The concept value does not exist, or
It is not a target ontology’s concept
Conditions An invalid ontology entity is used
Extensional spec. An invalid condition.
An invalid Property Bridge is one that has at
least one invalid argument according to the applied
service (Table 3).
Table 3: Property Bridge’s invalid conditions.
Invalid Argument Reason
Source Argument
It is not a source ontology entity, or
It doesn’t respect the service’s
interface
Target Concept idem
Conditions An invalid ontology entity is used
Extensional spec. An invalid condition.
By analyzing the SBO mapping document and
respective ontologies it is possible to identify the
invalid mapping entities.
3.3 Ontology Mapping Changes
The list of changes of ontology mapping largely
depends on the ontology mapping language.
Similar to other evolution approaches such as
object-oriented schema evolution proposed in
(Banerjee, et al., 1987) and ontology evolution
approaches (Klein, 2004) and (Stojanovic, 2004),
and according to the semantics of the SBO, a set of
ontology mapping changes were defined. In order to
improve the process, two abstraction levels of
ontology mapping changes were defined.
Elementary Ontology Mapping Changes.
Represent single changes in one SBO entity only.
Both additive and subtractive changes are
considered (Table 4). These changes form the
backbone of the mapping evolution system, in the
sense that they represent the modification at the
lowest level of complexity.
Table 4: Extract of elementary ontology mapping changes.
#id
Elementary Ontology Mapping Changes
1 AddConceptBridgeSourceConceptValue
2 RemoveConceptBridgeSourceConceptValue
3 AddConceptBridgeTargetConceptValue
4 RemoveConceptBridgeTargetConceptValue
5 RemoveHasBridge
6 AddHasBridge
Composite Ontology Mapping Changes.
Represent changes as the combination of two or
more elementary ontology mapping changes,
organized as one logical unit that should be executed
as a whole. Composite changes (Table 5) allow a
higher level of abstraction, representing a semantic
context of the mapping evolution process, because
they group together a set of meaningful and
complementary (elementary) changes, representing a
semantic change.
Table 5: Extract of composite ontology mapping changes.
Composite Ontology Mapping Changes
#id Elem.
Change
ChangeConceptBridgeSourceConceptValue 1,2
ChangeConceptBridgeTargetConceptValue 3,4
ChangePropertyBridgeHasBridge 5,6
3.4 Sorting the Invalid Entities List
Once the list of invalid entities in the mapping is
populated, it is necessary to sort it. The order in
which entities are processed is very important. In
fact, considering that entities are invalid because of
others, and because correcting one may correct
others, choosing the one that minimizes changes is
very important.
Figure
3 depicts the order in which the entities
are processed.
Because a Property Bridge is always defined in
the context of a Concept Bridge, Concept Bridges
must always be corrected first, so respective Proper-
A USER-DRIVEN AND A SEMANTIC-BASED ONTOLOGY MAPPING EVOLUTION APPROACH
217
Figure 3: Order of invalid entities to process.
ty Bridges are adapted accordingly. For example,
changing a Concept Bridge source concept argument
implies updating their Property Bridges’ arguments.
Additionally, for each Concept Bridge there is
also an established order of correction: first those
with invalid source and target concept values, then
those who have only one invalid concept value, and
later, those with invalid conditions.
Once all Concept Bridges are addressed and
respective Property Bridges’ arguments are updated,
invalid Property Bridges are addressed next.
By following this order, the process avoids
unnecessary computation and minimizes ontology
mapping changes, maximizing consistency and
maintaining the original semantics.
It is now possible to start correcting the invalid
entities. The next section describes a semi-automatic
user-driven approach, in which the user decides,
from a list prepared by the system, the corrective
action to execute. The Section 0 describes a
completely automatic approach based on the
semantics of the ontology evolution information.
4 USER-DRIVEN APPROACH
The correction process is potentially ambiguous
because several corrective mapping changes are
allowed for the same invalid entity (Table 6).
Table 6: Invalid mapping situations and corrective
mapping changes.
Invalid Mapping
Situation
Corrective mapping changes
Source AND Target
Concept
Remove Concept Bridge;
Source XOR Target
Concept
Remove Concept Bridge;
Change Concept Bridge Concept Value (to
super concept);
Source AND Target
Argument
Remove Property Bridge;
Source XOR Target
Argument
Remove Property Bridge;
Change Property Bridge Argument Value;
Extensional
Specification
Remove Concept/Property Bridge;
Remove Invalid Condition;
Generic Condition
Remove Concept/Property Bridge;
Remove Invalid Condition.
However, as previously mentioned, the execution
of a single ontology mapping change (e.g. remove a
Concept Bridge) may cause inconsistencies in others
mapping entities (e.g. subBridgeOf and hasBridge
relations). Thus, aside from the original change,
some other changes might need to be performed to
solve such problems.
In fact, two of the corrective mapping changes
identified in Table 6, potentially lead to ambiguous
situations (notice that the first invalid mapping
situation in Table 6 is not ambiguous):
Removing a Concept Bridge, triggers the
removal of the hasBridge relationships,
generating orphaned Property Bridges
(ambiguous situation). Additionally, triggering
the removal of subBridgeOf relationships lead
to orphaned Concept Bridges;
Changing Concept Bridge Concept Value may
cause inconsistencies in the Property Bridge’s
argument values, because the bridged
properties may not have the new concept
value in their domain. As a consequence, one
has to decide how to manage the Property
Bridges (ambiguous situation).
These ambiguities promote the concept of
mapping evolution strategy, similar to that used in
the scope of ontology evolution process. A mapping
evolution strategy is a pair between (i) one mapping
resolution point (ambiguous situation or a dilemma
that occurs during the ontology mapping process
and) and (ii) one elementary mapping evolution
strategy (a set of possible ways to solve a resolution
point).
Table 7 summarizes the resolution points and
respective corrective strategies.
Table 7: List of ontology mapping resolution points and
respective corrective strategies.
Resolution Point Elementary Mapping Strategy
Orphaned Property
Bridge
Don’t Propagate any Property Bridge
(Remove Property Bridges);
Propagate direct Property Bridges to
sub Concept Bridges;
Propagate (All) Property Bridges to
sub Concept Bridges.
Orphaned Concept
Bridge
Delete orphaned Concept Bridge;
Reconnect orphaned Concept Bridge
to parent;
Keep orphaned Concept Bridge.
For each mapping resolution point, one or many
elementary mapping strategies are possible.
The concept of Mapping Evolution Strategy
serves as parameterization of the system. In other
words, users choose how to solve each ambiguous
situation either on a local basis (i.e. the user decides
ICEIS 2009 - International Conference on Enterprise Information Systems
218
every time a resolution point occurs) or global (i.e.
the user decides to apply always a specific strategy
for every specific type of resolution point).
5 SEMANTIC-BASED
APPROACH
Consider the ontology evolution scenario depicted in
Figure 4, in which a “RemoveConcept" change has
been applied to the target concept T:C. While it is
obvious that the CB Concept Bridge has to evolve,
the changes to apply depend on the user decision,
and might not be the best solution.
Figure 4: Close-up of ontology evolution and mapping.
Because the ontology evolution strategies
capture the user’s intentions during that process, we
claim that they provide useful information for the
ontology mapping evolution process.
The goal is then to automatically identify the best
ontology mapping evolution strategy, based on the
ontology evolution information provided. This
should be understood not as a group of atomic
changes (e.g. RemoveConcept), but as well-
structured semantic information that fully captures
the ontology engineer intentions. To exploit such
information the concept of ontology evolution
strategy described in section 0 will be used.
Unfortunately, the chosen evolution strategy is not
explicitly registered in the log file, but only the
elementary changes, inter-related through the
“consequent” relation. For example, consider Figure
5 as a wider perspective of the scenario presented in
Figure 4, in which are represented not only the
original “RemoveConcept” change, but also the
deduced changes:
1 RemoveConcept(T:C)
2 RemoveSubClassOf(T:C,T:CSuper)
3 RemoveSubClassOf(T:CSub,T:C)
4 AddSubClassOf(T:CSub,Root)
5 RemoveDomain(T:C,P)
6 AddDomain(T:CSub,P)
7 AddDomain(T:CSub,PSuper)
The set of deduced changes provides evidence of
the application of a specific ontology evolution
strategy. From these, it is possible to generalise the
Figure 5: Example of ontology evolution and mapping.
necessary and sufficient conditions for the
identification of a specific ontology evolution
strategy.
Two resolution points may occur as a
consequence of the “RemoveConcept” change:
Orphaned Concept and Property Bridge Propagation.
The orphaned concept resolution point is addressed
by three elementary strategies: Reconnect to Root,
Reconnect to Parent and Delete Concept (Table 1).
The necessary and sufficient conditions (i.e. the
deduced changes) for the identification of the
“Reconnect Orphaned Concepts to Root” strategy
are defined in the next SWRL-like rule:
1 RemoveConcept(?TC)^
2 RemoveSubClassOf(?TC,?TCSuper)^
3 RemoveSubClassOf(?TCSub,?TC)^
4 AddSubClassOf(?TCSub,ROOT) ->
5 ReconnectOrphanedConceptToRoot(?TC).
The RemoveConcept change potentially gives
rise to “Orphaned Property” resolution points too,
which in turn are resolved by three elementary
resolution strategies. In the case when the
“Propagate All Properties” strategy is applied (as in
the case depicted in Figure 5), the
conditions/deduced changes are specified as:
1 ReconnectOrphanedConceptToRoot(?TC)^
2 Domain(?TC,?P)^
3 RemoveDomain(?TC,?P)^
4 AddDomain(?TCSub,?P)^
5 AddDomain(?TCSub,?PSuper) ->
6 PropagateAllProperties(?P).
For every invalid semantic bridge, the ontology
mapping evolution process tries to identify the
ontology evolution strategy adopted. When at least
one strategy is identified, the process applies a set of
ontology mapping changes that solve the invalid
mapping context.
A USER-DRIVEN AND A SEMANTIC-BASED ONTOLOGY MAPPING EVOLUTION APPROACH
219
The set of changes to apply was previously
identified as the best resolution process for the
ontology context considering the user intentions
applied at the ontology evolution. For every pair
<invalid mapping entity, ontology evolution
strategy> a set of ontology mapping changes are
defined. For example, the pair <invalid concept
bridge, ReconnectOrphanedConceptToRoot>, is
addressed by the following rule:
1 InvalidConceptBridge(?CB) ^
2 TargetConceptValue(?CB,?TC) ^
3 not(Concept(?TC)) ^
4 ReconnectOrphanedConceptToRoot(?TC)^
5 SourceConceptValue(?CB,?SC) ^
6 SubBridgeOf(?CB,?CBSuper) ^
7 SubBridgeOf(?CBSub,?CB) ^
8 TargetConceptValue(?CBSuper,?TCSuper)->
9 not(TargetConceptValue(?CB,?TC) ^
10 TargetConceptValue(?CB,?TCSuper) ^
11 not(SubBridgeOf(?CBSub,?CB)).
The predicate in line 1 identifies the invalid
concept bridge (?CB). Line 2 through 4 determines
that the target concept (?TC) value is missing and
that it has been removed and the “reconnect
orphaned concept to root” strategy has been applied
to its sub concepts. Line 5 through 8 instantiate
several variables (e.g. ?CBSuper is instantiated with
all super bridges of the invalid concept bridge). The
right hand side of the rule defines the mapping
evolution changes. Line 10 and 11 changes the
invalid concept bridge’s target concept value from
?TC to ?TCSuper. Line 11 removes the subBridgeOf
relationship between ?CBSub and ?CB.
Figure 6: Mapping scenario example evolved.
Similar expressions are defined for invalid
source concept and the remaining evolution
strategies.
Consider the mapping scenario depicted in
Figure 5. As a consequence of removing concept
T:C, both the “Reconnect Orphaned Concept To
Root” and the “Propagate All Properties” strategies
were applied. In response to these evolution
changes, the corresponding rules were executed,
giving rise to the mapping scenario depicted in
Figure 6.
6 EVALUATION
In respect to the user-based approach, no evaluation
has been made because when dealing with user-
defined strategies, the results are biased and should
not be evaluated by experience or test cases since the
results are subjective (Stojanovic, 2004).
In respect to the second approach, we claim that
because the mapping evolution changes captured in
the rules, mimic the changes adopted during the
ontology evolution process, they are (at least
indirectly) dependent on the user-defined strategies.
However, because there is a gap between the
ontology evolution intensions and those at the
ontology mapping evolution, we conducted some
tests based on two ontologies and four mappings. In
fact, despite there being no ontology evolution logs
currently available in the community, the research
team internally carried out the evolution of two
ontologies in the knowledge domain of R&D: (O1,
O1’), (O2, O2’) with 15-25 concepts (e.g.
researcher, publication) with 5-10 properties each.
The ontology evolution process adopted global
strategies. Two ontology engineers with upper
intermediate proficiency level in ontology mapping,
manually-mapped each ontology pair M1(O1,O2),
M2(O1-O2’), M3(O1’-O2), M4(O1’-O2’) for
reference, resulting in eight mapping documents.
The ontology mapping evolution process adopted
global strategies. We then applied the automatic
semantic-based ontology mapping evolution
approach to both M2 and to one of the M4
mappings. The obtained results were very similar to
the manually-specified mappings, but based on
differences, some minor refinements in the mapping
evolution rules were made. These refinements were
then included in the current version of the system.
Later, we applied the same approach to the
remainder mappings (i.e. both M3 and the other
M4). The results were again very similar but not
equal because users refined one of the mappings
ICEIS 2009 - International Conference on Enterprise Information Systems
220
with a local strategy. However, if that change was
not done, the results would be equal.
7 RELATED WORK
Handling ontology mapping documents is a relative
new research field, but similar problem has been
tackled in other contexts such as object oriented
(Banerjee, et al., 1987) or incremental view
maintenance (Ceri, et al., 1991). These approaches
present a taxonomy of changes, its semantics and a
set of deterministic rules. ToMAS (Velegrakis, et al.,
2003) is a tool for automatically detecting and
adapting invalid or inconsistent mappings due to
changes in either schemas or theirs constraints, even
if the changes did not make any of the mappings
syntactically incorrect. ToMAS also exploits
knowledge about user choices that is embodied in
the existing mappings. Further, while the changes
are not restricted to atomic type schema element, the
approach is schema centric and does not exploit the
semantics of the changes requested by the users.
Our approach adopts ideas from these
approaches but it goes a step further by exploiting
the evolution log, automating the process.
8 CONCLUSIONS
In this paper, the ontology mapping evolution
problem has been characterized in terms of goals,
inputs, constraints and outputs, and an abstract
ontology mapping evolution process has been
defined, comprised of two-iterative phases
respecting the removed and new entities in the
ontology, respectively.
Based on the semantics of SBO, a set of invalid
conditions and respective corrective changes were
identified. Yet, because different potential solutions
exist for the same invalid situation, a decision has to
be made. A user-driven approach has been
developed based on the structure and semantics of
the SBO. This approach guaranties that the mapping
document is corrected, but the user has to decide
which actions to take for every ambiguous situation.
Besides the list of possible corrective changes, no
further support is provided to the user.
Instead, the proposed semantic-based approach
automatically suggests the best corrective strategy,
based on the log information provided by the
ontology evolution process. The evaluation
performed showed that this is a valid and useful
approach, but further extensive evaluation has to be
carried out. However, this evaluation is not easy due
to the lack of logs.
A limitation of the semantic-based approach
concerns with the blind application of rules, based
on the necessary and sufficient conditions. In fact, it
is possible that an ontology evolution strategy has
been applied, even if only for some. Deciding
whether the strategy has been applied or not, is an
open issue. Additionally, because ontology mapping
largely depends on the ontology mapping language,
generalizing the proposed approaches to other
ontology mapping (or alignment) languages is a big
challenge for future work.
ACKNOWLEDGEMENTS
The work described in this document is partially
funded by the MCTES-FCT funded EDGAR project
(POSI/EIA/61307/2004) and the Coalesce project
(PTDC/EIA/74417/2006).
REFERENCES
Banerjee, Jay, et al. 1987; Semantics and implementation
of schema evolution in object-oriented databases.
SIGMOD Rec. Ca,US: 1987, Vol. 16, pp. 311-322.
Berners-Lee, Tim, Hendler, James e Lassila, Ora. 2001;
The Semantic Web. Scientific American. 2001, Vol.
284, pp. 34-43.
Ceri, Stefano e Widom, Jennifer. 1991; Deriving
production rules for incremental view maintenance. In
VLDB. 1991, pp. 577-589.
Euzenat, Jérôme and Shvaiko, Pavel. 2007. Ontology
Matching. Berlin: Springer, 2007.
IEEE. 1990; Standard Computer Dictionary: A
Compilation of IEEE Standard Computer Glossaries.
N.Y, U.S.: 1990.
Klein, Michel. 2004; Change Management for Distributed
Ontologies. 2004. PhD Thesis.
Plessers, P., De Troyer, O. 2005; Ontology Change
Detection using a Version. The Semantic Web –
ISWC. s.l. : Springer, 2005.
Rahm, Erhard and Bernstein, Philip A. 2001; A survey of
approaches to automatic schema matching. In VLDB.
s.l. : Springer, 2001.
Silva, Nuno. 2004. Multi-Dimensional Service-Oriented
Ontology Mapping. 2004. PhD Thesis.
Stojanovic, Ljiljana. 2004. Evolution, Methods and Tools
for Ontology. 2004. PhD Thesis.
Velegrakis, Yannis, Miller, Renée J. and Popa, Lucian.
2003. Mapping Adaptation under Evolving Schemas.
VLDB. Berlin: 2003.
A USER-DRIVEN AND A SEMANTIC-BASED ONTOLOGY MAPPING EVOLUTION APPROACH
221