AN ONTOLOGY CHANGE MANAGEMENT SYSTEM
An Experiment on a Health Care Case Study
Soumaya Slimani, Karim Baïna, Salah Baïna
ENSIAS, Mohammed V-Souissi University, Rabat, Morocco
Martin Henkel, Erik Perjons
Department of Computer and Systems Sciences, Stockholm University, Stockholm, Sweden
Keywords: Ontology Evolution, Semantic Service, Multi-agent System, Eye Specialist Ontology, Primary Health Care
Ontology.
Abstract: Numerous ontologies have been developed for life science domains. These ontologies are continuously
changing. Thus, it is becoming profitable to study and to manage these ontologies change in order to keep
all dependent ontologies and their related mappings consistent. The aim of this paper is to propose an agent
based approach enabling not only ontology and ontology mapping evolution analysis but also to manage
their changes. An experiment in health care illustrates the benefits of our approach. We apply our algorithm,
and implementation prototype p
2
OEManager to eye specialist ontology (ESO) and primary health care
ontology (PCO), and particularly, we use our ontology agent model, and prototype to manage some
significant changes in the ESO ontology.
1 INTRODUCTION
Ontologies become increasingly important in life
sciences. In electronic health care, the greater
problem is the heterogeneity of information systems.
Semantic interoperability of these heterogeneous
systems can be achieved through an agreement
between the underlying ontologies (e.g. RDF, OWL,
etc.). In the context of web services, several
standards were developed to describe web services
semantics (e.g. WSDL-S, WSMO, OWL-S, etc.).
Due to the rapid development of life science
research, ontologies evolve continuously, i.e., they
are frequently changing to incorporate new domain
knowledge into them. However, these changes, may
impact the correctness of future communication
using these ontologies because other services are not
aware of these changes. Hence, ontology mappings
(which allow services to interpret correctly
(translate) exchanged data) should be corrected.
Research in ontology evolution and change
management deal particularly with the versioning
and evolution of the same ontology, and do not
process the evolution of different ontologies,
interrelated by mappings and describing different
services. In this paper we propose an agent-based
algorithm and prototype managing the ontologies
evolution life cycle, as well as the evolution of
ontology-related mappings. Also, in a
comprehensive evaluation, we apply our algorithm
to eye specialist ontology (ESO) and primary health
care ontology (PCO), and, particularly, we use our
ontology agent model and p2OEManager prototype
to manage some significant changes in the ESO
ontology.
2 P
2
OEMANAGER DESIGN
AND IMPLEMENTATION
OVERVIEW
Fig. 1. shows the general architecture of
p2OEManager (which stands for peer to peer
Ontology Event Manager) upon 3 main components:
Ontology Manager, Ontology Mapping Manager,
and Ontology Agent Manager built on an open
Service Layer. Interactions between instances
p2OEManager peers (i.e. ontology change event
449
Slimani S., Baïna K., Baïna S., Henkel M. and Perjons E..
AN ONTOLOGY CHANGE MANAGEMENT SYSTEM - An Experiment on a Health Care Case Study.
DOI: 10.5220/0003105504490452
In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2010), pages 449-452
ISBN: 978-989-8425-29-4
Copyright
c
2010 SCITEPRESS (Science and Technology Publications, Lda.)
messages) are handled through an ontology change
communication channel. In fact, p2OEManager is
neutral regarding the service infrastructures.
Business messages between these services are
supported outside the scope of p2OEManager within
service marshalling/unmarshalling and security,
addressing, reliable messaging, routing, and
transport standard protocols.
Figure 1: p2OEManager Architecture.
As shown in Fig.1. , Service layer, which is outside
p2OEManager, represents services that are
described by service ontologies. These service
ontologies are defined by human designers using
ontology editors, and these service ontologies are
then monitored synchronously by p2OEManager.
Service dependencies are represented by service
ontology mappings defined by human designers too
using ontology mapping generators. Services
communicate and collaborate with each other within
service infrastructure based on ontology mappings
evolving in p2OEManager.
Our proposal is based on a combination of
ontologies and agents. We associate an agent to each
ontology. This agent is responsible of change
management and propagation of these changes to
other dependent ontologies. Table 1 summarizes the
four cases and the agent actions for each case.
For more formalisation details of our ontology agent
model, and algorithms implemented within our
p
2
OEManager Architecture, please refer to our
paper (Slimani and al., 2010).
Table 1: Dependent agent actions.
Syntactic Search
True False
Semantic
Search
True
-Updates the
related
mapping
-Updates the
mapping.
-Annotates the
corresponding
object by the added
object.
False
-Reformulates
the change
definition.
-Negotiates the
change
definition.
-Adds the new
object to the
ontology.
-Updates the
related mapping.
3 EVALUATION SCENARIO:
THE S:TERIKS HEALTH CARE
In order to illustrate the approach presented in this
paper a health-care case from the REMS project.
The main objective of the REMS project was to
develop a set of e-services that could be used to
create, manage and transfer health care referrals
between S:t Erik’s eye hospital specialist clinic and
primary care units. Having a set of e-services
available from different health care providers would
enable healthcare systems to be interconnected in
order to share information. To achieve this
integration, it is crucial for that the services share the
same set of concepts. Fig. 2 shows the value model
(Henkel and al., 2007) which depicts the main flow
of resources between the patient, the primary health
care and the eye specialist clinic. There are
important aspects that need to be considered when
designing e-services for the information exchange.
In the case of Swedish health services these should
follow standards on the international and national
levels. However, even given these standards it is still
plenty of room for interpretation of the concepts.
Furthermore, there is also a need to specialize the
concepts/models in order to cover specific details of
the Eye health care. Thus it is likely that two
organizations that follow the standardized
concepts/models will end up with two different
models that need to be kept synchronized if they are
to be able to exchange information.
Primary health care
unit
Patient
Eye specialist clinic
Initial opinion /
diagnosis
Ref erral to specialist
Patient fee
Information on
symtoms
Ref erral answer with information on
symptoms and diagnoses
Information on specialist
competencies
Performed
eye-treatment
Diagnosis
Information on ongoing
treatment
Referral,
describing the
health problems
Patient fee
Information on
symtoms
Figure 2: Actors and resource exchanges in the S:tEriks
health care case.
For this example, we focus on the interconnection of
the systems at the primary care units (PCU) and the
systems at the eye specialist clinics (ESC). Fig. 3
shows the basic ontology that is used on the PCU
systems. Fig. 4 shows the ontology that is used at the
ESC. These ontologies will be the basis for the
services that need to interconnect when exchanging
referral information. We have developed the
KEOD 2010 - International Conference on Knowledge Engineering and Ontology Development
450
Primary health care ontology (PCO) and the Eye
specialist ontology (ESO), using Protégé 3.4
(http://protege.stanford.edu).
Figure 3: Primary health care ontology (PCO).
Figure 4: Eye specialist ontology (ESO).
Common to both ontologies are the concepts of
PATIENT, ROLE, HEALTH PROBLEMS,
SYMPTOMS and REFERAL. We have generated a
mapping using Prompt which generate an OWL file
for the mapping
1
(http://protege.stanford.edu/plugins/prompt/prompt.h
tml). For the sake of describing our approach we
provide in the next section four examples of changes
4 ALGORITHM APPLICATION:
PROCESS OF CHANGES
As input of the process of using the Ontology Agent
Model, we have the Primary health care ontology
(PCO), the Eye specialist ontology (ESO), and the
mapping between PCO and ESO (MEP). The
process of using the Ontology Agent Model can be
presented in 3 steps: (1) We have paramerezed the
Ontology Agent Model by integreting PCO, ESO,
and MEP URIs. We have obtained two Ontology
Agent: one for the Primary health care ontology
(PCO) and another for the Eye specialist ontology
(ESO), (2) run the PCO agent and the ESO agent,
(3) Process changes.
The application of the algorithm in this scenario
begins on the Initiator side (ESO agent). First,
changes are listed as follow:
changeSet = <
C1=<ADD :Class :EYE_REFERAL:
subclassOf REFERAL>,
C2=<ADD:Class:HEALTH_CARE_ACTIVITY:sub
classOf ACT>
C3=<ADD:Class:PARTICIPATION:subclassOf
Thing>
C4=<ADD:Class:ORGANIZATIONAL_UNIT
:subclassOf Thing>
C5=<ADD:ObjectProperty:ORG_CG:<Domain:
ORANIZATIONAL_UNIT, Rang:CARE_GIVERS>>
>
Then changes are classified according to their
relationship with the mapping. All changes are sent
to PCO agent which process for each change the
correspondents actions as follow:
For C1. The syntactic and semantic search return
false. So, PCO update the mapping by adding the
following correspondences: mapped (
ESO. C
1
, PCO.C
1
)
For C2. The syntactic search return false, but
semantic search will return true, because
HEALTH_CARE_ACTIVITY can be mapped to the
existing concept of HEALTH_CARE_EVENTS in
the PCO. Thus the PCO agent (1) Annotate the
concept HEALTH_CARE_EVENTS concept by
adding Rdfs:seeAlso HEALTH_CARE_ACTIVITY
annotation, (2) Update mapping by adding the
following correspondence :
mapped(
HEALTH_CARE_EVENTS,HEALTH_CARE_ACTI
VITY
)
For C3. The syntactic and the semantic search
return true. So, the PCO agent updates the mapping
by adding the following correspondence:
mapped(
ESO.PARTICIPATION, PCO.PARTICIPATION)
For C4 and C5. The concept
ORGANIZATIONAL_UNIT is added with an
association to CARE_GIVERS in the ESO.
ORGANIZATIONAL_UNIT matches syntactically
to ORGANIZATIONAL_UNIT in the PCO, but this
match is semantically incorrect since only
PUBLIC_O can have ORGANIZATIONAL_UNIT
in the PCO. So, the PCO agent calculates adefinition
for ORGANIZATIONAL_UNIT as follow:
<<Class:PCO. ORGANIZATIONAL_UNIT:
subclass of Thing>
<ObjectProperty: PCO.OU_HAS :
Domain: PCO.RGANIZATIONAL_UNIT,
Rang :PCO.PUBLIC_O >>
AN ONTOLOGY CHANGE MANAGEMENT SYSTEM - An Experiment on a Health Care Case Study
451
PCO Agent sends this definition to ESO agent,
which calculates a new definition:
<<Class:PCO.ORGANIZATIONAL_UNIT:
subclass of Thing>
<ObjectProperty:PCO.OU_HAS:
Domain: PCO.RGANIZATIONAL_UNIT,
Rang:PCO.PUBLIC_O>
<Class:ESO.ORGANIZATIONAL_UNIT:
subclass of Thing>
<ObjectProperty: ESO.OU_HAS :
Domain:ESO.ORGANIZATIONAL_UNIT,
Rang:ESO.CARE_GIVERS>>
The following correspondence introduces a conflict
in the definition of CARE_GIVERS:
mapped (ESO.CARE_GIVERS, PCO.PUBLIC_O)
Indeed, as we have already in the mapping that
ESO.PUBLIC_O corresponds to PCO.PUBLIC_O
and ESO.PUBLIC_O is a subclass of
CARE_GIVERS, this can be a source of errors. So,
the agent sends an alert message to the user. User
can modify the mapping and the ontology or validate
any of the definitions contained in the negotiation
exchange. In this example, our algorithm allows
services in the Eye Specialist Clinic and in the
Primary Care Provider to be aware of evolution in
other services. Agents take the necessary decisions
to maintain a reliable exchange of data between
these entities.
5 RELATED WORKS
The ontology evolution and change management has
been addressed by many researches. (Klein and al.,
2001). (Klein, 2004), investigated the versioning of
ontologies, (Plessers and al., 2007) define Change
Definition Language (CDL), (Djedidi and al., 2009)
a patterns-based ontology evolution approach ,
(Zablith, 2009) a Framework for ontology evolution
and (Hartung and al., 2008). But we propose an
evolution model for multi-ontology system when
ontologies are different and not only for instance.
When ontologies are considered as an ontology
instances, the source ontology has the semantics of
the dependent ontologies (because they are instances
of the source ontology). So, ontology evolution will
require applying the same changes on the dependent
ontologies. Note also that in the case of different
ontologies, mapping between ontologies must be
managed in parallel.
6 CONCLUSIONS AND FUTURE
WORK
Interconnecting services is a complex task. A part of
the complexity comes from the need to have a
common, shared view of the information. The
approach proposed in this paper is suitable for
ontology evolution management in distributed and
heterogeneous environments. The aim is to provide a
flexible way to partially automate the process of
ontology evolution management. The approach
consists of software agents that represent each of the
services involved, and their respective ontologies
and related mappings. To illustrate our approach
we applied it to a health care case. The case study
highlighted some of the main benefits of the
approach. The approach could be extended and
improved in several ways. First, there is a need to
further analyze the implications that changes have to
the logic of the running software services.
Furthermore, there is a need to extend the ontology
agent model to include different types of changes in
our algorithms (removal, etc...).
REFERENCES
Slimani,S.,Baïna,S.,Baïna,K., “Agent-based Architecture
for Service Ontology evolution Management”,
SEKE’10, San Francisco Bay, USA 2010.
Henkel, M., Johannesson, P., Perjons, E., Zdravkovi, J.,
"Value and Goal Driven Design of E-Services", The
IEEE International Conference on e-Business
Engineering (ICEBE'07), Hong Kong, China, October
24-26.
Klein, M., Fensel, 2001. D. Ontology versioning on the
Semantic Web. In: Proc. Int. Semantic Web Working
Symposium (SWWS).
Klein, M., 2004. Change Management for Distributed
Ontologies. PhD thesis, Vrije Universiteit Amsterdam.
Plessers, P., Troyer, O., and Casteleyn, S., 2007.
Understanding ontology evolution : A change
detection approach, Web Semantics : Science,
Services and Agents, on the World Wide Web 5 39-
49.
Djedidi, R., Aufaure, MA., 2010. ONTO-EVOAL an
Ontology Evolution Approach Guided by Pattern
Modeling and Quality Evaluation. FoIKS.
Zablith, F. 2009. Evolva : A Comprehensive Approach to
Ontology Evolution”. In Proceedings of the 6th
European Semantic Web Conference PhD Symposium
LNCS 5554. eds. L. Aroyo et al., Springer-Verlag,
Berlin, Heidelberg, pages 944-948, ESWC.
Hartung, M., Kirsten, T., Rahm, E., 2008. Analyzing the
evolution of life science ontologies and mappings. In:
Proc. of DILS.
KEOD 2010 - International Conference on Knowledge Engineering and Ontology Development
452