ADAPTATION DRIVEN CHANGE MANAGEMENT
Lai Xu, Paul de Vrieze
SAP Research Switzerland, Blumenbergplatz 9, CH-9000 St.Gallen, Switzerland
Athman Bouguettaya
CSIRO ICT Centre, Canberra, ACT, Australia
Jian Yang
Department of Computing, Macquarie University, Sydney NSW 2109, Australia
Keywords: Change Management, User Adaptive Systems.
Abstract: In solving the global obesity epidemic the realization has come that it is mainly an issue of behavioral
change. Benefits have been shown in personalizing the information that induces these changes. At the same
time obesity is a heavily researched field in which new discoveries are made every day. This brings the need
for maintainability of these personalized information channels. In this paper we propose architecture for a
maintainable system for the provision of tailored documents for obesity patients.
1 INTRODUCTION
Obesity is a global epidemic that has serious impact
on the occurrence of non-transmittable chronic
diseases. The treatment of obesity is understood to
be found in behavioral change. The transfer of
information forms a significant part of the efforts to
change the dietary behavior and activity levels of
obesity patients. It is understood that different
obesity patients react differently to different
information dependent on their personal properties.
Using computers to generate tailored brochures
based on user characteristics has been tried by
various research groups and found to have a positive
health outcome (Brug et al., 1996, Brug et al., 2003,
Kreuter et al., 1999).
As obesity is a chronic disease, and behavior
change has to be actively monitored and redirected
to be retained. As such it is important that the means
used to achieve this will remain available through
this lifelong process. Obesity however is also an area
where a lot of research is being done. This means
that the system must be able to evolve with the
knowledge it is built upon without loosing the
knowledge gained from previous interactions with
the patient.
Developing such a system, we face some
frequent changes. For example, new or
complementing research of obesity treatments
becomes available now and then. The personal
conditions of each patient may change during the
process. Trust in the system or health professionals
from each patient will be gained or lost. Personal
characteristics are influenced by the experiences of
the weight management process. Furthermore, the
change of adaptation strategies of the system should
also be taken into account. Knowing, adapting to
and managing changes are thus important aspects for
the design of the system.
There are three types of changes in the weight
management system. First, the changes of personal
conditions, personal characteristics, and trust in the
system or health professionals from the patient are
naturally processed by the user adaptive system.
Second, new research results on obesity treatment
can be seen as new knowledge. In the area of
knowledge management the subject of how to
detect, handle changes of knowledge and organize
information in a better way have been broadly
investigated. Third, the changes that cannot be
handled by either user adaptive systems or
knowledge management systems. These changes,
288
Xu L., de Vrieze P., Bouguettaya A. and Yang J. (2009).
ADAPTATION DRIVEN CHANGE MANAGEMENT.
In Proceedings of the International Conference on Health Informatics, pages 288-293
DOI: 10.5220/0001550102880293
Copyright
c
 SciTePress

such as adaptation dynamics, changes and
integration, are the research focus of this paper. The
majority of research in user adaptive system and
knowledge management system is focused on
construction issues. Coping with the changes and
providing maintenance facilities however require a
different approach.
Change mining in the context of decision tree
classification for real-life application is studies in
(Liu et al.,2000). Their primary goal is to know what
is changing and how it has changed in order to
provide the right products and services to suit the
change market needs. In our system, the goal of
managing changes of the users' interactions with the
systems and changes of knowledge is not only to
predict changes but also to improve the efficiency of
the system.
Paper (Berendt et al., 2002) discusses the
possibilities to combine the two research areas
Semantic Web and Web Mining. The idea is to
improve the results of Web Mining by exploiting the
new semantic structures in the Web, and to make the
user of the Web Mining for building up the Semantic
Web. The ideas inspire us to define semantics for the
log information and using the log information to
enhance ontologies and usage of system.
In the areas of knowledge management and
ontology evolution, Heflin (Heflin, 2001) argues the
needs of evolution of ontologies on the Web. A new
formal definition of ontologies is provided for the
use in dynamic, distributed environments. The way
of performing the change is however not dealt.
Ontology evolution can be treated as a part of the
ontology versioning mechanism that is analyzed in
(Klein and Fensel, 2001). An overview of causes
and consequences of the changes in the ontology has
been provided. A detailed analysis of the effect of
specific changes on the interpretation of data is
missed.
We provide an ontology-based approach for
adaptation models and discuss issues of adaptation
models in section 2. Ontology changes in the system
are argued in section 3. Section 4 provides The
conceptual architecture of the weight management
system. Finally section 5 summarizes concluding
remarks.
2 NEEDS OF ONTOLOGY IN
USER ADAPTIVE SYSTEMS
In user adaptive systems the system maintains a user
model. This user model is updated to reflect actions
of the user. When the system then needs knowledge
about the user to tailor itself it does so by inferring
from the user model the answers needed. What the
user model looks like, how the system updates the
user model and which inference is used to get the
needed answers is described by the adaptation model.
While an adaptation model, or a meta model for
a user model is sufficient to describe the behavior of
a user adaptive system at a particular point in time, it
can not be used to relate with different systems, or
even different adaptation models for the same system.
Ontologies allow one to express the relation
between two adaptation models. As user models are
instantiation of an adaptation model, the ontology
can thus allow the transition of a user model for the
old adaptation model to a user model for the new
adaptation model.
diff = am’ – am
um’ = apply(diff, out, um)
For this and other purposes we need another
layer on top of the adaptation model. This ontology
layer defines the meanings of the concepts used in
the adaptation model. The ontology, for example,
describes for the possible user properties what they
mean and what type they have.
Ontologies are also essential in supporting
change of adaptation models. As adaptation models
get updated the user models need to be updated
accordingly. As there can be many user models this
updating should not need to be manual.
As an example of how ontology can help support
an adaptation model a containing two properties:
weight and BMI (Body Mass Index). The body mass
index is defined as the individual's body weight
divided by the square of their height. As this system
is used it shows to be inconvenient to update BMI as
well as weight. Given that the height of adults is
more or less constant, it would be better to store
height instead and calculate the body mass index on
demand.
To perform this change we first verify that the
ontology contains the concepts weight, height and
BMI and their relation. Then we create, based on our
original adaptation model a a new adaptation model
a'. In this new adaptation model the BMI user
property is replaced by a height property and a BMI
function. As the relation between the concepts is
defined by the ontology the BMI function can be
created automatically. Given a and a' we now need a
function f(u
a
)=u
a
that updates user models for a to
user models for a'. This function again can be
determined automatically from the ontology to
calculate height from BMI and weight and store this
height instead of BMI.
ADAPTATION DRIVEN CHANGE MANAGEMENT
289

Ontologies can further be used for adaptation
model composition. With the rise of ubiquitous
computing multiple sources of user information
become more and more probable. In our weight
management system the patient information can be
built from different devices such as mobile phones,
PDAs and so on. In this respect it is sensible to
extend the model to allow multiple applications to
share a user model.
3 ONTOLOGY CHANGES
Our weight management information system
provides weight control strategies to individual user
cases. The system maintains knowledge of diet,
exercise and so on. The tailored advice is generated
according to different weight control stages and
individual conditions. In the context of the weight
management information system, the purposes of
using ontologies are:
 semantic annotation of knowledge fragments,
 describing relationships between knowledge
fragments,
 selecting and assembling demanded knowledge
fragments according to the ontology, and
 sketching adaptation model elements.
The first three purposes are normal ontology uses in
knowledge management systems. The last use is a
specific use for our weight management information
system. Contrary to one's intuition in our ontology-
based application ontologies need to be timely
adapted to changed requirements from environments
and users.
3.1 Types of Changes
Ontology change can be classified from two
dimensions. One dimension can be viewed as
implicit changes and explicit changes whereas
another dimension specifies the origin of changes,
namely internal or external changes. The external
requirements of changes come from the
environment, while the internal changes can be
required by the system, such as the usage of
ontologies in the system. Table 1 summarises all
changes expected from different perspectives.
Table 1: Changes of Environment and System.
Changes Internal /
System
External / Environment
Implicit Concept creep External sources change
usage of ontology.
Explicit Fixes, turning,
refinement
New knowledge
First, we will show an example of internal
implicit changes. To show this we distinguish the
written ontology from the effective ontology as it is
used by the system. The existence of an effective
ontology can be shown in analogy with a natural
language, such as the living English language, where
the dictionary needs to be updated regularly to
reflect changes in the meanings of words. As an
ontology is used by humans, such subtle change of
meaning can also occur in ontologies, changing the
effective ontology. Thereby causing the effective
ontology to no longer be equal to the written
ontology.
Implicit change can also happen externally. If a
system incorporates information from external
sources, their effective ontologies can change. The
ontology then needs to be update to reflect this
external change. For example, in a web service
environment, suppose one hotel chain decides that
the provision of a mini-bar will be standard in its
two-star rooms while it is not a standard requirement
for a two-star hotel. The two-star hotel can update
the room specification independently. As a customer
who is using a hotel search system, querying about
hotel rooms with a mini-bar. The customer may not
be able to find the two-star hotel with a mini-bar,
because the ontology in the hotel search system
specifies only standard contents of two-star hotels.
Being able to overcome such problems, the ontology
in the hotel search system should be added a new
subconcept or property for the concept of two-star
hotel.
Explicit internally driven change can be seen in
the maintenance of the system and ontology, such as
fixing, turning and refining ontology. For instance,
from user feedback of a hotel search system, the
system administrator knows that customers complain
about the inability to check whether ``Heineken''
beer is served by the hotel. Therefore, subconcept or
property ``Heineken'' of concept ``Beer'' should be
added. In our weight management system, when a
new adaptation strategy is found, deploying a new
adaptation strategy is also explicit internal change.
Finally, explicit, externally driven change is for
example caused by the need to add new knowledge
to the system which needs new concepts added to
the ontology. For example, introducing a new
concept, such as eight-star hotel rooms in a hotel
search system.
3.2 Needs of Ontology Changes
While it is interesting to see where change
originates, one should not neglect looking at the
need for change. If one understands the needs for
change, one knows what to look for so that the
needed changes are made. If the underlying ontology
HEALTHINF 2009 - International Conference on Health Informatics
290

is not up-to-date, annotated and assembled
knowledge fragments can be inconsistent, redundant
or incomplete. The reliability, accuracy,
effectiveness and adaptivity of the weight
management system thus decrease significantly.
Like ontology change, the need for ontology
change can be regarded from the two axes of
implicit versus explicit and internal versus external.
There is no one to one mapping between ontology
change and need for ontology change though. The
summary of needs for ontology changes is presented
in Table 2.
Table 2: Needs for Ontology Changes.
Needs for
ontology
changes
Internal External
Implicit Lack of query
focus, precision or
recall;
Unused concepts for
the current
ontology.
Abandonment of
concepts etc.;
Inexistent
concepts for the
current ontology
Explicit Incompatibility of
new concepts with
ontology;
Inefficient
treatments.
Out-of-date
information;
Adopting new
user models.
To address the needs for ontology change,
assessment of an ontology-based system is needed.
From an information retrieval perspective, the
criteria are the improvement of precision and recall
of the system. From adaptive system, adaptivity is
another standard.
An implicit, internally motivated need for change
could for example be that the system has a bad
precision (on query), or that a concept is never used.
It requires one to actively look at what is wrong.
An implicit external need could be the
abandonment of concepts by an external source, or
access to new knowledge was the ontology to be
updated. For this reason it is useful to monitor the
actual usage of term and identifies new terms such
as what is ``eight-star hotel room''.
An explicit internally motivated need for change
would for example be to updated the ontology to be
compatible with new knowledge. Like a new
treatment does not comply with the used ontology to
annotate.
Finally, an example of an explicit external need
for change would be the change of a law that applies
to the system. An example of a change of a law is
updating four-digit zip codes into five digits.
4 ARCHITECTURE
The management of an ontology-based user adaptive
system is complex and essential for a sustainable and
extensible information system dealing with changes
in user's need as well as environment. We present
architecture (in Figure 1) of a user adaptive system,
which can handle changes in the adaptation model.
In order to efficiently retrieve advice, the content
of the knowledge resource (authorised document, cf.
1 in Figure 1) is annotated in terms of the underlying
domain ontology (10). The annotation process can
be supported by an annotation tool (2). Annotations
can be explicitly added to the knowledge resources.
The annotated knowledge sources are stored in a
knowledge repository (3).
The knowledge is used in the response handler.
The document composition engine (4) is responsible
for assembling the needed document based. The
response planner (5) determines which document the
document composition engine should compose.
The adaptation component is the part of the
system that is responsible for modeling the user, and
making this knowledge available to the rest of the
system (de Vrieze, 2006). An appropriate way to
look at the adaptation component would be as a
virtual copy of the user that does not mind
answering lots of questions. As such the question
handler (6) takes the questions from the rest of the
system, and based on the user model (7) determines
the best answer. The user model stores the
knowledge on the user, and is maintained by the user
behavior analyser (8).
The event dispatcher (9) is responsible for
feeding the user actions, as events, to the various
interested parts of the system. Events are sent to the
response planner such that the system may respond.
Events sent to the user behavior analyser allow it to
take the events into account in the user model.
Events sent to the log allow for later, off-line,
analysis of usage of the system that forms an
important basis for the recognition of the need for
evolution of the system.
Ontologies (10) store ontologies used for annotating
knowledge and ontologies used for changing or
integrating adaptation models. Ontology usage is
monitored in the monitoring component (13). It
results in a set of recommendations for changes to
the annotations and/or domain ontologies. These
recommendation sets are processed in the ontology
evolution component (12). The recommendations
can still be accepted or rejected by the knowledge
engineer. The ontology evolution component needs
to ensure the consistency of the system (incl.
ontology and all dependent artifacts) after resolving
changes.
ADAPTATION DRIVEN CHANGE MANAGEMENT
291

Figure 1: Conceptual Architecture.
HEALTHINF 2009 - International Conference on Health Informatics
292

Information whether and how the knowledge
repository reflects the needs of end-users can be
obtained by analyzing the user's interaction with the
system. For example which subjects are looked at,
how many results are delivered, which documents
are found and so on. This data is captured in the log
file (14). The log file is also processed in the
``Change need discovery & usage monitoring''
component (13). In contrast to the analysis of
annotation, the log-file analysis might result in the
requirements to update the knowledge repository.
The bugs and suggestions reported by the final users
of the system are stored in the bug list (15).
The changes derived from the user's behavior are
application-oriented and, after applying these
changes, the system should improve its
performances. The task of the validation component
(11) is to detect changes that decrease system
performance and to react appropriately. E.g. by
rolling back the changes involved, restoring the
original state of the whole system (the ontology,
annotations and the knowledge repository) in the
case of decreasing performance.
The ontology editor and ontology coverage
mapper (16) is an interface for ontology
management within the system. For example for the
coverage mapper it integrates with the knowledge as
stored in the system.
5 CONCLUSIONS
We derived a novel approach for dealing with
weight management. Specifically the dynamic
nature of the user behavior and the obesity related
research. The approach is based on a user-adaptive
system. In particular, we consider changes from
both user behavior perspectives as well as changes
of the (knowledge) environment. In order to allow
the system to handle the evolution during long-term
usage, the design of the system takes into account
the sustainable management of change.
The benefits of the proposed approach are
manifold: (i) the system can handle changing user
behavior and build up a long term trust relationship,
The system further allows the integration of different
adaptation models. (ii) The system will handle the
change of the environment. The evolution process
enables continuous system improvement by semi-
automatic discovery of changes. (iii) The system
provides a sustainable information system. The
validation component helps the maintainer in better
understanding of effects of each change providing
detailed insight into each change being performed.
ACKNOWLEDGEMENTS
Part of work from Lai Xu and Paul de Vrieze has
been performed at the CSIRO ICT Centre, Australia.
REFERENCES
Moore, R., Lopes, J., 1999. Paper templates. In
TEMPLATE’06, 1st International Conference on
Template Production. INSTICC Press.
Smith, J., 1998. The book, The publishing company.
London, 2
nd
edition.
Berendt, B., Hotho, A., and Stumme, G. 2002. Towards
Semantic Web Mining. In Proceedings of the First
international Semantic Web Conference on the
Semantic Web (June 09 - 12, 2002). I. Horrocks and J.
A. Hendler, Eds. Lecture Notes In Computer Science,
vol. 2342. Springer-Verlag, London, 264-278.
Heflin, J. 2001. Towards the Semantic Web: Knowledge
Representation in a Dynamic, Distributed
Environment. PhD thesis, University of Maryland,
College Park.
Klein, M. and Fensel, D. 2001. Ontology Versioning on
the Semantic Web. Proceedings of the International
Semantic Web Working Symposium (SWWS), pages
75–91.
Lerner, B. S. and Habermann A. N. 1990. Beyond schema
evolution to database reorganization. In
OOPSLA/ECOOP ’90: Proceedings of the European
conference on object-oriented programming on
Object-oriented programming systems, languages, and
applications, pages 67–76, New York, NY, USA.
Liu, B. Hsu, W. Han, H.-S. and Xia, Y. 2000 . Mining
changes for real-life applications. In DaWaK 2000:
Proceedings of the Second International Conference
on Data Warehousing and Knowledge Discovery,
pages 337–346, London, UK. Springer-Verlag.
Mann, W. and Thompson, S. 1987. Rhetorical Structure
Theory: A Framework for the Analysis of Texts.
University of Southern California, Information
Sciences Institute.
National institutes of Health, National Heart, Lung, and
Blood Institute, NHLBI Obesity Education Initiative,
and North American Association for the Study of
Obesity. 2008. The practical guide: Identification,
evaluation, and treatment of overweight and obesity in
adults.
http://www.nhlbi.nih.gov/guidelines/obesity/prctgd_c.
pdf. Last accessed Feb 4, 2008.
NHS. Clinical knowledge summaries.
http://www.cks.library.nhs.uk/.
de Vrieze, P. 2006 Fundaments of adaptive
personalisation. PhD thesis, Radboud University
Nijmegen. ISBN-13: 978-90-9021113-8.
Wing, R. and Phelan, S. 2005. Long-term weight loss
maintenance. American Journal of Clinical Nutrition,
82(1 (S)):222–225.
ADAPTATION DRIVEN CHANGE MANAGEMENT
293