Authoring Educational Topic Maps: Can We Make It Easier?
Darina Dicheva and Christo Dichev
Winston-Salem State University, Winston-Salem, N.C. USA
{dichevad, dichevc}@wssu.ed
Abstract
2. Difficulties in authoring educational TM
In this paper we examine some problems related to
capturing the structure and the topic name space of
learning content in the context of Topic Map authoring.
We demonstrate that the use of traditional course
taxonomies as ontological resources is problematic.
Based on these findings and on the results of a locally
conducted study, we propose an empirically justified
minimal ontology for Topic Maps-based e-Learning. The
proposed minimal set contains five relation types and is
implemented in the TM4L Editor to support authors that
experience difficulties in articulating and naming
relationships.
The authors of educational Topic Maps are typically
instructors who create maps for the courses they teach. In
order to find out what are the major difficulties they face
we conducted a study in which seven TM were created
with the TM4L Editor by different authors. The study
showed that authors generally didn’t have problems in
selecting appropriate learning content and resources.
However, they had difficulties in structuring the content
and defining and naming relationships between concepts
(topics). In the next sections we discuss these problems.
2.1. Classifying learning content
1. Introduction
Current Web-based educational practices indicate that
courseware authors’ ability to gather and generate
information exceeds their ability to organize, manage,
and effectively use it. Ontologies are a key technology
emerging to facilitate Web information processing by
supporting semantic structuring, annotation, indexing,
and search. Ontologies allow organization of learning
material around components of semantically annotated
topics. This enables ontology-based educational systems
to do efficient semantic querying and navigation through
the learning content. We have developed an authoring
tool, the TM4L (Topic Maps for Learning) Editor [1][2],
which enables the development of ontology-aware
courseware based on the new Semantic Web technology
Topic Maps [3].
In the Topic Map (TM) paradigm an ontology is an
accurate description of the essential entities and relations
which are found in the modeled domain, and can be
represented as a set of topics linked by associations.
Therefore the Topic Maps technology is well suited for
structuring learning material around subject ontologies.
We are currently experimenting with the TM4L
Editor by creating educational Topic Maps for different
university courses. In this paper we discuss the main
problems authors face when creating ontology-based
courseware and propose strategies for overcoming some
of the TM authoring problems.
In a typical courseware structure, learning content is
laid out in a tree-like structure of course units (lectures,
sections, subsections etc), an approach adopted from the
traditional textbooks organization. In general, the names
of the course units have some relationships based on the
author’s notion of classification. It is appealing to think
that the course units can be organized based on a
hierarchy of concepts (terms), like a taxonomy. However,
this is not true in general. Frequently the concept
structure used to organize the learning content is not a
proper hierarchy, and the concepts naming the sub-units
of the learning material do not represent more
specialized content of their “parents”. For example, in
the Prolog book of Sterling and Shapiro the topic “Lists”
is under “Recursive programming” while in Bratko’s
book it is under “Lists, Operators, Arithmetic”. This
shows that the order in learning content classifications is
often subjective and arbitrary and could easily be
reversed. However, if the order could be altered and still
make sense then that topic structure doesn’t represent a
true hierarchy. Apparently this subjective approach
encourages ad hoc concepts organizations. Moreover,
authors’ intuition about where to place a unit sometimes
is inconsistent with the broadly adopted structure. This
may result in putting a unit in an unexpected for learners
place. When several authors are involved, it is even
harder to keep a consistent organization of the
instructional material. An additional complication is that
both the names and organization of concepts are subject
of change over time. Even the ACM Computing
Classification System has been changed several times
since its first publication.
2.2. Identifying topics
The structure of the learning content usually reflects
the author’s concept of systematization. The titles (topic
names) and their relations depend on authors’ knowledge
and goals at the time of creation. Being subjective they
are of variable quality and with uncontrolled terms.
There are no rules to limit the authors to use specific
information for describing content. Uncontrolled
vocabularies make it easy to record information but shift
the load of interpretation to the users.
Among the principal problems with identification and
naming topics are:
•
•
The titles (topic names) are not necessarily unique.
Generally there are no agreed terms for all topics.
• The titles may not be informative enough.
• The titles fail to group related materials together
in a more or less standard way.
Further practical questions include: What is the
relationship of title subject and the actual content? How
does one phrase come to present a multitude of subjects?
These questions reflect problems related to browsing and
searching for relevant resources. An example in this
context is the use of different titles to represent the same
topic. In such cases, a search for '”Pattern Matching” will
not pick up items indexed with the term “Unification”.
On the other hand, the title is an identifying label by
which we refer to the subject. The extension of titles’
original use in the new context of e-learning requires a
stable, durable topic identification system.
hierarchy (in TM4L representing “class-subclass”
relations). The author of the Topic Map “Basic Counting
Principles”, for example, used only the default
hierarchical type beside the relation “created by” for
expressing all relations between the defined topics.
The parent-child topic classification reflects the titlesubtitle tradition established by the conventional
textbook organization and the created TM-based learning
material mirrors the context in which authors used to see
the composing items. The problem though is that an
organized collection of learning items often represents a
kind of contextually related topics which is difficult to
translate into conventional hierarchical structures.
Another problem comes from the fact that in a TM
browser only one hierarchical relation type is usually
displayed as a tree. Therefore only the topics connected
with relationships of that type would be displayed in the
tree. All other topics will be seen not linked to it. Thus,
the authors cannot actually see all created topics
arranged nicely in a tree-like structure. An attempt to
“fix” this problem brought the author of the “Number
Systems” TM to define two different relationships
instead of one between many topics – “instance-of”
which she felt is the proper one and “class-subclass” just
to display the topics in the domain term (ontology)
hierarchy.
When a relationship between two topics is not
‘hierarchical’, the problem is even worse since the author
has to decide also how to name it. Unlike concepts,
which are generally named by terms from the subject
domain, there are no established/agreed names for
relations. The latter are usually named by common
language words and there is a variety of words that can
express their meaning. The choice of different words by
different authors poses though a serious problem for the
reusability and exchangeability of created courseware.
2.3. Articulating and naming relations
3. Support for authoring educational TM
As we already mentioned, TM authors have
difficulties in deciding what type of relationships to use
and how to name them. Authors generally try to follow
the content structure of the used textbook, which might
not be taxonomy: any given topic may represent a
subclass, an instance, a property, a “See also”
relationship, etc. Typical collections of learning content
are a somewhat incoherent combination of taxonomies,
partonomies, and other (possibly unnamed) schemes. The
authors shared that often they would encounter a relation
between two concepts of the kind “more general - less
general” but could not determine its exact type, not to
speak about a name. When the relation between the
topics was ‘hierarchical’, they would place the new topic
as a child of an earlier created (parent) topic in the topics
Educational Topic Maps authors are generally
untrained in information classification and work in lack
of controlled vocabularies and support from ontology
analysts. We decided to support the TM4L users in two
ways: (1) Help them reuse existing established
classifications (combined with controlled vocabularies);
(2) Make available some predefined relations to them.
Here we propose an empirically justified minimal
ontology for expressing contextual relationships in
educational Topic Maps. By predefining the minimal set
of relation types we enable authors that experience
difficulties in articulating and naming relationships to use
the predefined general relationships. Our model contains
two layers: standard and author’s layer. The standard
layer contains five predefined generic relationships: the
classic “superclass-subclass” and “class-instance”, one
general “hierarchical” and two “horizontal” relationships.
The author’s layer might contain specialization of the
general relations defined by the author.
3.1. Modeling general hierarchical relations
Our goal was to define a generic relation that captures
the current practice of organizing learning material in a
structure of topics similar to a table of content.
Following the work on SKOS [4], we suggest extending
the Topic Map model for the needs of the educational
TM with a new relation, which we call “super-sub”. This
relation carries weak semantics used to express the fact
that one topic is more general than another. This implies
that the related topics can be arranged into a hierarchy,
without being too strict about the exact semantics of the
relationship. The proposed relation is a generalization of
“part-whole” and can be used to characterize associations
with asymmetric roles assigned to two role players, such
as X is-part-of Y, X is-based-on Y, etc. “Super-sub” is
not a version of “superclass-subclass” or “class-instance”
relationships: it is less restrictive than these two. The
most informative properties of our “super-sub” relation
derive from transitivity used also in type hierarchies such
as “instance-of” and “class-subclass”. However it is more
general and defined as an asymmetric, transitive relation
satisfying the condition if (a R b) then not (b R a). The
insight was to provide a generic, extensible hierarchical
relation, intended to serve as a type of a family of
relationships with more specific meaning where the
semantics of the new relations is inherited from the
“super-sub” relation. The “super-sub” relationship
captures in generic sense hierarchical relationships
within the modeled domain, such as “part-whole”,
“section-subsection”, “folder-subfolder”, “based-on”,
“supported-by”, etc. Thus, instead of classifying concepts
such as “facts”, “rules” and “queries” as instances or
subclasses of “Prolog”, authors can use the predefined
relation “super-sub”.
3.1. Modeling “horizontal” relations
If semantically similar resources are scattered among
distinct topics, a simple hierarchical browsing will not
find them, while the TM technology allows finding
references that are horizontally related and unexpected.
We propose two horizontal relationships between topics:
“relevant to” and “mentioned by”. The first one is
introduced to capture relations with symmetric roles
assigned to two role players, such as “co-refers”, “is
similar to”, “is synonym of”, “is of same level of
complexity”, etc. It can be used to represent mapping
between equivalent topics. The relation satisfies the
condition if (a R b) then (b R a).
The second relation “mentioned by” represents the
family of asymmetric, not necessary transitive relations.
It is used to express the fact that two topics are related,
but the relationship can not be used to build hierarchies.
The relation “mentioned by” is intended to capture in a
generic sense asymmetric relations of the type X
mentions Y, X refers-to Y, X discusses Y, X is used by
Y, X is created by Y etc. The key relations in the
pragmatic sense are the binary relations. The basic
intuition is that the five proposed relations (“superclasssubclass”, “class-instance”, “super-sub”, “relevant to”
and “mentioned by”) represent a sufficient basis of
generic relations for creating educational Topic Maps.
They can be used as a generic grouping of concepts and
resources that might be difficult to articulate relieving the
authors from creating dummy or incorrect relation types.
4. Conclusion
We have created an authoring tool, the TM4L Editor
that supports the development of standards-based
ontology-aware online learning materials. In this paper
we discuss some problems authors face when creating
ontology-based courseware with TM4L and a strategy for
overcoming one of them. We propose an empirically
justified minimal relation set for TM based e-learning in
an attempt to support authors that experience difficulties
in articulating and naming ontology relationships.
5. Acknowledgement
This material is based upon work supported by the
National Science Foundation under Grant No. DUE0333069 “NSDL: Towards Reusable and Shareable
Courseware: Topic Maps-Based Digital Libraries.”
6. References
[1] Dicheva, D., Dichev, C. “A Framework for Concept-Based
Digital Course Libraries”, J. of Interactive Learning Research,
15(4), 2004, pp. 347-364.
[2] Dicheva D., Dichev C., Sun, Y., Nao, S. “Authoring Topic
Maps-based Digital Course Libraries”, SW-EL@AH 2004, The
Netherlands, 2004, 331-337.
[3] Biezunski, M., Bryan, M., Newcomb, S.: ISO/IEC
13250:2000
Topic
Maps:
Information
Technology,
www.y12.doe.gov/sgml/sc34/document/0129.pdf.
[4]
SKOS
Core
Vocabulary
http://www.w3.org/2004/02/skos/core.
Specification,