ACTIVE LEARNING BY PERSONALIZATION
Lessons Learnt from Research in Conceptual Content Management
Hans-Werner Sehring, Sebastian Bossung, Joachim W. Schmidt
Hamburg University of Science and Technology, STS - Software Systems Institute
Harburger Schlossstraße 20, 21071 Hamburg, Germany
Keywords:
Content-based and Context-based Learning, Research Systems, Knowledge Management, Personalization.
Abstract:
In this paper we argue that the potential of e-learning systems for active and autonomous knowledge acquisi-
tion and use can be substantially enhanced by exploiting the personalization capabilities of concept-oriented
content management.
Content and implementation personalization are essential services provided by research-oriented content man-
agement systems. Personalized concept understanding as well as concept-oriented content acquisition and
use are key to most learning situations. Therefore, learning and research systems should cooperate with the
common goal of exchanging and sharing content and enabling processes which transparently span system
boundaries. In this paper, we discuss how personalization can be used to enhance both, autonomous learning
activities and research-oriented workflows.
Personalization and content coupling technologies are at the heart of one of our operational web-based appli-
cation systems, the Warburg Electronic Library. This system is successfully used in a number of research as
well as learning projects, during which advantages of joint research and learning systems have been identified.
1 INTRODUCTION
Essential goals of research-oriented content manage-
ment systems – in short, research systems – and e-
learning systems overlap. Both system classes sup-
port autonomous user activities such as acquisition
and exploration of concepts as well as creation, en-
richment, publication, and sharing of content. As a
consequence, a close coupling of both system classes
is considered as mutually beneficial. Furthermore,
key requirements are present in both, most promi-
nently personalization. Research support systems
focus on the representation of facts by content. It is
important, however, that they do not impose a spe-
cific interpretation of content on users. As there is
generally no agreed-upon single interpretation, such
systems need to allow users to work with personal in-
terpretations. Moreover, associating content with a
fixed sets of a priori categories is inappropriate be-
cause every researcher needs to focus on aspects spe-
cific to the work at hand.
In addition to content organization, storage and re-
trieval, research systems have to support three types
of research workflows: (1) personalization of given
content and structures, (2) sharing of personalized
content amongst users, and (3) making the views of
users available to the general public. According to
our findings many of these requirements can be met
by means of personalization.
E-learning systems generally support learners in
finding and studying content appropriate for their
learning needs; content is packaged in learning units
stored in e-learning systems. Such learning units
can often be parameterized to better suit the learners’
previous knowledge or their preferences with respect
to learning styles, (Guttomsen Sch
¨
ar and Krueger,
2000). An integral part of learning, however, is
putting the heard into practice to gain a thorough un-
derstanding of the subject area at hand, e.g., (El Sad-
dik et al., 2001). To allow learners to work seamlessly
during their entire learning activity, systems also have
to support autonomous environments for active learn-
ing. We have found that active participation of users
is also enabled by means of personalization.
In addition, there is a need for content linking
and exchange between research and learning systems:
Many of the research results of today evolve into
learning materials of tomorrow.
We have gained many of the insights presented here
through our operational system, the Warburg Elec-
496
Sehring H., Bossung S. and W. Schmidt J. (2005).
ACTIVE LEARNING BY PERSONALIZATION - Lessons Learnt from Research in Conceptual Content Management.
In Proceedings of the First International Conference on Web Information Systems and Technologies, pages 496-503
DOI: 10.5220/0001235404960503
Copyright
c
 SciTePress

tronic Library (WEL). The WEL research system em-
phasizes the support of research by flexibility in both
structure and content, as well as for the respective
processes. Both can be personalized to match the re-
searcher’s intentions. Since 2000, the WEL system is
successfully used in a number of research projects.
Because of the similar requirements we have also
been able to employ it as a learning system, demon-
strating the advantages of joint research and learning
systems. User groups include researchers in art his-
tory, school teachers, and university students.
The paper is organized as follows: We begin with
details about personalization (section 2) and show in
section 3 and section 5, respectively, how it is em-
ployed in research and learning systems. Section 4
gives an account of our experiences with an imple-
mentation of the concepts presented in this paper. In
section 6, we provide a closer look at the coupling
technology for research and learning systems. We
conclude with a summary and outlook in section 7.
Related work is discussed where appropriate.
2 PERSONALIZATION
By personalization we mean the ability of a system to
adapt to the individual needs of each user. Personal-
ization can be broken down into two aspects: schema
openness and system dynamics. An open schema is
one that users can change on-the-fly and at any time,
thus guaranteeing best correspondence with their in-
formation needs. Information systems are dynamic if
their implementation follows any on-the-fly modifica-
tion of a schema dynamically.
Such personalization happens on two levels: con-
tent and structure personalization. The former refers
to the ability of a system to let users change content
according to their needs or opinions, the latter enables
users to change the schema of the content. (Rossi
et al., 2001) refer to content personalization as infor-
mation personalization.
Note that this personalization is more general than
what one is used to from contemporary web applica-
tions, where personalization usually covers presenta-
tional aspects only. (Koch and Rossi, 2002) also share
this broader definition of personalization.
Another important aspect of personalization is the
ability of users to re-integrate their changes with the
content base of a larger community (fig. 1). This al-
lows for iterative enhancement of the content. Per-
sonalization is used to achieve a number of goals:
(1) Adaptation of content to personal views or in-
tents, (2) securing of privacy, and (3) handling of large
amounts of data by filtering out what is not of inter-
est, (Nieder
´
ee, 2002). Also see (Riecken, 2000) on
personalization.
system-wide
content
group
content
personal
content
personalize
integrate
personalize
integrate
Figure 1: Personalization allows local change and re-
integration of content.
2.1 Content Personalization
Research work is explorative and often subjective. It
follows open and dynamic processes (see above, or
(Sehring, 2004) for more details) in which every user
has an own – possibly changing – view on the content.
This personal view is reflected by individual modifi-
cations of the content. Users can – at least for some
time – deviate from the opinion of the community by
changing a piece of content. Personalization allows
them to modify existing content, but those changes
will only be visible to the user who made them. This
approach is therefore called content personalization.
Content personalization can be used to explore the im-
plications of a hypothesis.
Note that presentation personalization (e.g., chang-
ing the background color) is a special case of content
personalization as configuration data can be treated
just like ordinary content.
2.2 Structure Personalization
Content personalization allows users to modify con-
tent according to their current needs without interfer-
ing with others. However, in many situations this is
not sufficient, as it can handle uniform content only.
Structure personalization comprises two aspects:
(1) Changing the structure of content by means of
creating variants of the content’s schema as well as
(2) (re)categorizing content with the option of creat-
ing new categories. As with content personalization,
there need to be means that allow users to re-integrate
their changes with the community. Structure person-
alization is a dynamic application of schema evolu-
tion, e.g. (Banerjee et al., 1987).
For example, often users create a new attribute, to
capture newly found aspects. In addition, they might
also want to modify the categorization of content.
2.3 Implementing Personalization
Personalization as required by research and learning
systems is not covered by contemporary information
ACTIVE LEARNING BY PERSONALIZATION - Lessons Learnt from Research in Conceptual Content Management
497

server module
assets
data
adapted assets
assets
proxy assets
remote assets
unified view
view 1 view 2
XML documents
assets
mediation module
distribution module
mapping module
client module
Figure 2: Modules interface with each other in a layered
architecture.
systems (ISs). Since ISs are usually based on database
technology they share its typical constraints, the most
crucial being that databases rely on one static schema.
The schema is adapted to users’ information needs by
means of views or the application layer of a multi-tier
application.
A static schema prevents dynamic processes, as all
parts of an IS –data, application, andpresentation lay-
ers – refer to that schema. Changing it leads to mod-
ification of the whole application. Since this requires
manual changes to the software, it cannot be done by
end-users. The same is true for view changes. Hence
the demand for openness is violated.
After first experiments with monolithic multi-
schema applications which enable openness and dy-
namics by means of complex application code we
now favor a new approach to IS construction. This
approach is based on a conceptual modeling language
for content and concept schemata as described in
(Schmidt and Sehring, 2003). This language intro-
duces the notion of an asset as an indivisible union
of perceivable content representing an entity and a
set of expressions describing it abstractly. Conceptual
expressions include values of fundamental properties
of entities, relationships between (otherwise indepen-
dent) entities, and constraints governing rules and reg-
ulations which entities follow.
From definitions given in the asset definition lan-
guage, by end-users, open dynamic systems are gen-
erated by a technology that resembles model-driven
architecture approaches, (Sehring and Schmidt,
2004). It consists of a model compiler and a mod-
ularized architecture for asset-based systems, which
can be used to create research and learning systems.
Users can express their information needs in the
asset definition language without regard to imple-
mentation constraints. The model compiler estab-
lishes openness by generating software systems meet-
ing those requirements.
m
client
: client module
for database with
schema for  M
m
map
: mapping
module  M M’
m’
client
: client module
for database with
schema for  M’
m
med
: mediation module with schema M’
delegating to m’
client
and m
map
DB with
schema
for M’
DB with
schema
for M
Figure 3: Sample configuration of a personalization from
schema M to M
′
.
In our approach the architecture of the generated
research and learning systems achieves personaliza-
tion by the ability to enable dynamic systems evolu-
tion. For this it allows the open redefinition of assets
and the dynamic invocation of the model compiler.
A system consists of a set of components reflecting
one schema each. These are broken down into mod-
ules, which are the output of the model compiler. For
our current purposes we identified five kinds of mod-
ules (see fig. 2); each module of a component is of
one of the five kinds. By introducing module kinds
we achieve a separation of concerns which makes ad-
dition and substitution of modules possible.
Personalized content is stored in third party sys-
tems, databases in most cases. Mapping asset sche-
mata to schemata of such systems is done by client
modules.
The functionality of a component is defined by
a module configuration. Therefore a central build-
ing block of the architecture of most applications is
the mediator architecture presented by (Wiederhold,
1992). In our approach it is implemented by mod-
ules of two kinds, namely mediation and mapping
modules. Fig. 3 shows an example of a configura-
tion which enables personalization. The component
is accessed via mediation module m
med
. It distributes
requests according to a personalization strategy, e.g.,
to create and manipulate content objects in a user’s
private storage (via m
′
client
) while retrieving content
from both the public (via m
client
) and the private
store. The mapping module m
map
converts requests
formulated according to the personalized schema M
′
to ones according to a global schema M .
As can be seen in fig. 3 the public component
(m
client
and the corresponding database) is integrated
into a personalized component unchanged. This is
the key to dynamic personalization, here changing the
content structure from M to M
′
.
By use of distribution modules, components spread
out over severalsystems. They are accessed via server
modules using standard protocols.
WEBIST 2005 - E-LEARNING
498

3 RESEARCH SYSTEMS
The support of research activities is esentially a con-
tent management task, because information is re-
trieved as a basis to research and new content is cre-
ated as its result. Software which supports research
has – in extension of common content management
systems – to provide content in a cooperative and per-
sonalizable way. The software has to meet require-
ments which go beyond those of common content
management systems and will be discussed in sec-
tion 3.1.
Very roughly, doing research can be broken down
into two parts, which are shown in figure 4. The first
is working with existing knowledge and facts to de-
rive new findings from them. This is shown in the
inner cycle. In a research support system, this typi-
cally involves customizing existing content and then
changing it to reflect new ideas.
The outer cycle in figure 4 describes the interaction
with the scientific community. After a number of iter-
ations in the inner cycle, researchers will publish their
results. This serves two purposes: Making their find-
ings known to the world as well as receiving feedback
from the world.
An integral part of research is the discovery and
construction of new concepts. These make it possi-
ble to put the new findings into relation with existing
knowledge, thus aiding the understanding of new dis-
coveries. To this end, it is important to note that re-
taining existing concepts and content structures will
be hindering at best. The old structures were devised
in previous research iterations and can only be used
to store information the corresponding structures. Es-
pecially the categories will fit the new findings only
partially.
Customize
Personal
findings
Publish
Knowledge
Enrich
Research
content
Figure 4: Research cycle showing personal actions of the
researcher as well as interactions with the community.
3.1 Research System Requirements
In extension to typical demands to research systems
(Yao, 2003), we identify additional requirements.
Alltogether, researchers need to be able to:
• create new structures (i.e., both new concepts and
new data structures to store findings) as well as new
content, which represents the actual findings,
• categorize existing as well as new content accord-
ing to their needs, which might involve changing
the categorization of existing content or finding ev-
idence (content) for existing categories,
• manage co-existing structures of concepts to han-
dle personalized as well as public research results,
• enter a review process (e.g., by peer groups) which
ensures consistency of the new findings with exist-
ing ones, and to
• republish their findings to the outside world once
the research is complete.
A research system needs to have an additional prop-
erty: It needs to ensure that intermediary steps in the
research are only visible to the researcher and not to
the outside world. Intermediary results hardly ever
make sense to anybody but their creator. In many
fields of research (e.g., in social studies) there are also
privacy concerns.
Ordinary information systems are ill-suited as re-
search support systems as they generally do not meet
many of the requirements. Shared data is the com-
munication paradigm of information systems (Goldin
and Thalheim, 2000). These systems therefore allow
neither changes in structure nor in content without
making the changes visible to all users at the same
time. Even if categorization is handled, on-demand
re-categorization is usually not supported.
3.2 Personalization in Research
Systems
By means of personalization one can create a system
that meets the above requirements.
First of all, it allows the implementation of the in-
ner cycle of fig. 4: Researchers can take existing con-
tent and change both its value as well as its schema.
They can also restructure content on a larger scope,
e.g., by changing or adding categorizations. This
process can be repeated until the personal findings are
in a consistent state.
Then researchers can choose to publish the find-
ings to a larger community. This can happen in sev-
eral steps, for example first to other members of the
research group, and only later to the general public.
Here again support is given by personalization, be-
cause it also includes an upward path that allows the
ACTIVE LEARNING BY PERSONALIZATION - Lessons Learnt from Research in Conceptual Content Management
499

Figure 5: A view of the running WEL system.
reintegration of personalized content into the content
that was originally personalized from, see figure 1.
Personalization is thus the outstanding feature that
enables the conduct of research in the system.
4 A RESEARCH SYSTEM IN
E-LEARNING SCENARIOS
The Warburg Electronic Library (WEL for short
1
)
has contributed much to the authors’ insights into re-
search and learning support systems, (Schmidt et al.,
2001). It is publicly available as a web-based system
since 2000. It is and has been used in several research
and learning projects by a current total of over 700
users world-wide.
4.1 Research
The WEL software was originally developed for the
project “Regent Base” which was carried out in co-
operation with art historians from the research in-
stitute for Political Iconography at the University of
Hamburg. Analyzing the demands of the human-
ities we soon discovered that they cannot express
their findings in conventional databases since subjec-
tive views dominate those disciplines. Therefore the
WEL has been designed as a personalizable digital
1
Also see the web site: http://www.welib.de
library which follows the paradigm of reference li-
braries (Nieder
´
ee, 2002).
The personalization methodology of the WEL is
based on the understanding that many organizational
paradigms for information, for example those of pub-
lic libraries, reference libraries, and card indices, all
share the same fundamental principles. Besides the
amount of content, they differ primarily in the content
schemata. The organizational structure is determined
by the size – and thus the broadness of interests – of
the addressed community.
The users of the WEL are organized in groups,
which are laid out in hierarchical manner to allow
researchers to capture their concerns in projects and
subprojects. Group membership determines the con-
tent which is visible to the user. In combination with
transparent personalization of content, this allows for
individual research as well as collaborative work in
groups. Groups exist for a range of projects with of-
ten different but sometimes overlapping content.
4.2 Learning
The WEL system has been used for e-learning pur-
poses since 1998. Several seminars have been car-
ried out with students of art history. A common
task in such a seminar is to structure content from
a given subject area, e.g., a regent, a geographical
region, or an epoch. By relating content and find-
ing subject terms describing the area at hand, stu-
dents gather information on the political situation, a
regent’s goals, or the ways of mass communication
by art. Finally, students give a presentation on their
findings and write a text about them. Using the WEL
system we experimented with employing a research
system to systematically support such seminar work.
We used the chance to try different interaction pat-
terns of users in different roles. The general approach
of using the WEL’s personalization in the seminars is
depicted in fig. 6. This workflow meets the demands
from (Maurer and Lennon, 1995).
At the beginning of each course the instructors pre-
pare material for student groups working on different
topics (“customize”). Using personalization they se-
lect WEL content and make it available to the groups
as a starting point. In this first interaction pattern
teachers communicate with learners using the WEL.
Students use the prepared content to start a personal
card index for their topic. They structure additional
content, which they either find in digital repositories
available through the WEL or add by digitizing mate-
rial from books or catalogues (“personalize”).
While finding, structuring, and amending content
(“(re-)structure and improve”) the students are able
to communicate with each other. By making content
available to members of other groups on a peer-to-
peer basis they are able to discuss results and bring
WEBIST 2005 - E-LEARNING
500

“Virtue”
card index
work group 3
“Mantua and the Gonzaga ”
reference library
seminar 1999/2000
“Index of Political Iconography ”
public library
Martin Warnke
personalize
(re-) structure
and improve
publish as
multimedia
artifact
customize
Feedback:
- notification
- negotiation
- acceptance
- ...
Figure 6: An example of the learning cycle with system
support: the seminar process.
out interrelations of topics. This second interaction
pattern allows students to jointly work on their topics.
It trains their ability to work in teams and to partici-
pate in scientific discourse (Alavi et al., 2002).
While working on a topic, students can present ma-
terial to their instructors since both are WEL users.
Furthermore, instructors can monitor the progress of
students and guide them. Both ways of communica-
tion form the third category of interaction patterns:
learner-to-teacher communication.
When they are done with the acquisition of con-
tent, students demonstrate their findings by giving a
talk. Normally they do this by taking attendants on
one linear path through their structure, while showing
content where appropriate. Using the WEL, they are
able to visualize the complex structures they created,
resulting in more in-depth discussions.
Finally, students prepare a medial presentation of
their findings. All past seminars created a web page
for this purpose, as indicated in fig. 7.
In a current WEL-project, the “Hamburg Media In-
dex”, the WEL system is taken to school teaching
to explore the remaining fourth interaction pattern,
namely teacher-teacher interaction. As school curric-
ula are currently being reformed, teachers have to col-
lect and create new learning material. With the help
of the WEL, teachers can jointly work on new mater-
ial and discuss it with colleagues. Individual teachers
can also offer content to others. Personalization gives
them control over the publication of their materials.
5 LEARNING SYSTEM
REQUIREMENTS
As is witnessed by many taught courses, absorbing
content and applying the learnt are essential parts
of learning, see for example (Clancy, 1995). Con-
structionists interpret learning as the construction of
knowledge and not its absorption, (Schroeder, 2002).
Artist
Virtue
artist
image text
publish
Minerva
Garden of
Virtue
image
classifies
aggregates
text
Mantegna
painted by
Figure 7: Multimedia publication of content.
Practical application facilitates the lasting storage of
information by guaranteeing the incorporation of the
appropriate association into the existing cognitive net-
work. By means of this association the learnt can later
be applied to different, more complex problems.
5.1 Learning Activities
Fig. 8 depicts this active way of learning. There are
two levels, which differ in closeness to the learner as
well as in speed of iteration. The inner cycle of pur-
poseful action, getting results on the action, and in-
corporating those into the mental model focuses on
active learning. It is usually carried out by one learner
alone. The outer circle describes the interaction with
others, learners as well as teachers.
Typical e-learning systems support a passive way
of learning: There are usually a number of ways
to present lessons to learners (Guttomsen Sch
¨
ar and
Krueger, 2000). Interaction of learner and system is
frequently limited to formal testing.
Unfortunately, the abilities of e-learning systems to
allow the learner to solve problems inside the system
itself are limited. Enabling this is crucial to support
active learning, which is believed to be important to
allow learners to gain a thorough understanding.
5.2 Personalization in E-learning
Ideally consumption of content in e-learning systems
is augmented by making the content parameterizable.
That is, users are allowed to view the same content in
different forms. To this end, the content is parameter-
ized and users can tweak its presentation according to
ACTIVE LEARNING BY PERSONALIZATION - Lessons Learnt from Research in Conceptual Content Management
501

Purposeful
action
Mental
model
Discourse
Shared
understanding
Perceiving
Results
Figure 8: Learning cycle from the learner’s point of view.
Inspired by (Allert et al., 2004), but content is different.
their needs. Typical dimensions of parameterization
include the language of presentation, the number of
examples, or the degree of formalism.
While content personalization can be used to im-
plement parameterization, personalization itself cov-
ers a much broader scope. Parameterization mainly
applies to teaching, while personalization covers all
aspects of the learning circle in figure 8. It supports
discourse through exchange of personalized content
and structures with a limited group of peer learners.
Most importantly, users are enabled to take action in
the system itself and learn through the results of their
action. They can recombine artifacts to solve learn-
ing problems. This applies to the subject under study
(which is represented in the system in one form or an-
other through content) as well as learning objects (the
course lessons which deal with the subject).
This allows a new approach to e-learning in which
learners can structurally rework or even extend the
subject matter. For example, a lesson could start with
a given partial content structure that the learners are
to complete. In doing so, they apply what they pre-
viously learned. Thinning out the content is again
achieved through personalization (the left-out pieces
are of course not deleted globally), as the whole sys-
tem is likely to provide a much higher level of detail
in the given area, for example to also support more
advanced learners.
6 COUPLED SYSTEMS
Active participation of the learners is an important
part of the learning process. Practical experience has
shown that this in general can be enabled by means of
personalization.
While the need for personalization is shared by
both learning and research systems, the feature – con-
tent as well as structure personalization – is used in
different ways in each kind of system. It is a criti-
cal part in the learning process to enable active work
by learners, to allow learners to share knowledge in
groups, and to enable teachers to cooperatively de-
velop learning content. In research systems person-
alization allows users to formulate hypotheses, de-
velop them in isolation, and finally share them with
the community in a controlled manner.
Research and learning systems can be linked on
a technical level since they share the same under-
standing of personalization. Therefore research and
learning processes can be coupled, allowing the ex-
change of research results and learning content. Joint
processes can transparently span system boundaries.
Combining research and learning processes allows for
various synergies not possible with unrelated systems.
Findings obtained through research systems will
eventually make it into teaching. Joint processes sup-
port such a migration by keeping a link from course
material to research content. This link is lost when
content is copied manually from one system to an-
other, discarding information about, e.g., authorship
and data provenance, see (Buneman et al., 2001).
Occasionally, advanced learning content con-
tributes to research. For example, the students’ results
from the seminars reported on above caused parts of
the project’s content to be refined. Similar results can
stem from Master’s or PhD thesis work which link
back to research content. Tracking the life-cycle of
content from a research system into a learning system
and back is not feasible without system support.
7 SUMMARY
The success of research and learning systems relies
heavily on their support for active processes in open
and dynamic environments. To our findings, such sys-
tems are largely based on personalization functional-
ity for content, structure, and the processes attached.
The required personalization exceeds the capabilities
of traditional information systems. Implementation
and application experience shows that advanced per-
sonalization can be successfully employed for both
research and learning purposes.
To extend the approach presented here towards
Intelligent Tutoring Systems (ITSs), models of the
learners and teaching strategies are required in addi-
tion to the model of a domain. The asset definition
language allows the open formulation of these models
comparable to, e.g., user models capturing expertise
in research systems by keeping track of their inter-
WEBIST 2005 - E-LEARNING
502

ests, contributions, and ratings by others. By means
of ITSs, models of learners can be used to give feed-
back and to dynamically adjust learning modules to
the learners’ abilities. How to make use of a set of
learner descriptions to determine typical obstacles is
currently being researched.
Future work will extend the principles discussed
in this paper essentially in two dimensions. First,
we seek to extend the interaction patterns of the
WEL system by designing and evaluating the teacher-
teacher communication discussed above and group-
oriented student-student interactions. Second, we see
a need for interoperability with other e-learning sys-
tems through standardized e-learning formats such as
LOM, SCORM and the like. Furthermore, the use of
exchange formats based on XML will also allow gen-
eral sharing of any content. In our approach XML
schemas are dynamically generated out of user de-
fined asset models. A first prototype implementation
of an XML schema generator exists.
In a varity of application projects we are seeking
additional learning and research scenarios to acquire
further user demands that will challenge the architec-
tural decisions for our personalizable conceptual con-
tent management systems.
ACKNOWLEDGEMENTS
We would like to thank our colleagues from the Art
History Department of Hamburg University, first of
all Prof. Dr. Martin Warnke, for the extraordinary pa-
tience of a highly experienced user community. Fur-
thermore, we thank Hamburg County for its contin-
uous financial support of the Warburg Electronic Li-
brary Project (WEL grant) and Deutsche Forschungs
Gemeinschaft (DFG) for its WEL-related grant on
“Geschichte der Kunstgeschichte im Nationalsozial-
ismus”.
REFERENCES
Alavi, M., Marakas, G. M., and Yoo, Y. (2002). A com-
parative study of distributed learning environments on
learning outcomes. Info. Sys. Research, 13(4):404–
415.
Allert, H., Richter, C., and Nejdl, W. (2004). Lifelong learn-
ing and second-order learning objects. British Journal
of Educational Technology.
Banerjee, J., Kim, W., Kim, H.-J., and Korth, H. F. (1987).
Semantics and implementation of schema evolution in
object-oriented databases. In Proc. ACM SIGMOD,
pages 311–322. ACM Press.
Buneman, P., Khanna, S., and Tan, W.-C. (2001). Why and
Where: A Characterization of Data Provenance. In
Proc. of the 8th ICDT 2001, volume 1973 of LNCS,
pages 316–330. Springer-Verlag.
Clancy, W. (1995). A Tutorial on situated learning. In Proc.
Int. Conf. on Computers and Education.
El Saddik, A., Fischer, S., and Steinmetz, R. (2001).
Reusable Multimedia Content in Web-Based Learning
Systems. IEEE MultiMedia, 8(3):30–38.
Goldin, D. Q. and Thalheim, S. S. B. (2000). IS = DBS
+ Interaction: Towards Principles of Information Sys-
tem Design. In Proc. ER 2000, 19th Int. Conf., volume
1920 of LNCS, pages 140–153. Springer-Verlag.
Guttomsen Sch
¨
ar, S. and Krueger, H. (2000). Learning
Technologies with Multimedia. IEEE Multimedia,
7(3):40–51.
Koch, N. and Rossi, G. (2002). Patterns for adaptive web
applications. In Proc. 7th European Conference on
Pattern Languages of Programs.
Maurer, H. and Lennon, J. (1995). Digital Libraries as
Learning and Teaching Support. Journal of Univer-
sal Computer Science, 1(11):719–727.
Nieder
´
ee, C. (2002). Personalisierung, Kooperation und
Evolution in digitalen Bibliotheken. PhD thesis, Ham-
burg University of Science and Technology.
Riecken, D. (2000). Introduction: personalized views of
personalization. CACM, 43(8):26–28.
Rossi, G., Schwabe, D., and Guimares, R. (2001). Design-
ing Personalized Web Applications. In Proceedings
of the tenth international conference on World Wide
Web, pages 275–284. ACM Press.
Schmidt, J. W. and Sehring, H.-W. (2003). Conceptual Con-
tent Modeling and Management. In Proc. Perspectives
of System Informatics.
Schmidt, J. W., Sehring, H.-W., Skusa, M., and Wien-
berg, A. (2001). Subject-Oriented Work: Lessons
Learned from an Interdisciplinary Content Manage-
ment Project. In et al, A. C., editor, 5th East Eu-
ropean Conference, ADBIS 2001, volume 2151 of
LNCS, pages 3–26. Springer.
Schroeder, U. (2002). Meta-Learning Functionality in
eLearning Systems. In Proc. Int. Conf. on Advances
in Infrastructure for Electronic Business, Education,
Science, and Medicine on the Internet.
Sehring, H.-W. (2004). Konzeptorientierte Inhaltsverwal-
tung: Modell, Systemarchitektur und Prototypen. PhD
thesis, Hamburg Univ. of Science and Technology.
Sehring, H.-W. and Schmidt, J. W. (2004). Beyond Data-
bases: An Asset Language for Conceptual Content
Management. In Proc. 8th East European Conference
ADBIS 2004.
Wiederhold, G. (1992). Mediators in the Architecture of
Future Information Systems. IEEE Comp., 25:38–49.
Yao, Y. Y. (2003). A Framework for Web-based Research
Support Systems. In 27th International Computer
Software and Applications Conference.
ACTIVE LEARNING BY PERSONALIZATION - Lessons Learnt from Research in Conceptual Content Management
503