EXTENDING EXISTING E-LEARNING PLATFORMS TO

SUPPORT AUTOMATIC CONVERSION OF LEARNING CONTENT

TO SCORM
Polyxeni Arapi, Nektarios Moumoutzis, Yiannis Maragoudakis, Stavros Christodoulakis
Laboratory of Distributed Multimedia Information Systems and Applications
Technical University of Crete (MUSIC/TUC)
Chania, Greece
{xenia, nektar, imarag, stavros}@ced.tuc.gr

Abstract
A critical issue in eLearning domain today is the preservation of existing learning material and user groups and
learning communities already set up in the context of operational eLearning environments as the eLearing
environments are being transformed to support the new interoperability specifications. This issue has tremendous
economic impact. In this paper we address this issue in the context of a pre-existing eLearning platforms implemented
on top of relational database management systems. The assumed architecture is based on two assumptions: (1) the
pre-existing eLearning platform is implemented on top of a relational database management system which is used to
store all the relevant information and (2) the model to be supported comes in the form of an XML schema. The
prototype implementation refers to web-based eLearning platform (Distant Learning Center) based on a solid
pedagogical framework (Virtual University Paradigm). The interoperability model adopted is SCORM 1.2. The
proposed approach is based on a multi-tier architecture and makes use of software components that allow for the
automatic creation of Package Interchange Files (PIF) that could the be uploaded to any SCORM conformant
Learning Management System.

Keywords
SCORM, interoperability, eLearning

1. INTRODUCTION
The issue of education has historically received much attention in periods of social transition. All
intellectuals aiming at changing the society or some aspects of it have also stressed the role of education. It
is thus evident that the current era of transforming our post-modern societies in the so called "information
society" pays much attention to education and training as a means to ensure that the necessary
competencies and skills are mastered by all the population. Nevertheless the transition phase that we are
experiencing today is even more dependent on education and training because of the inherent emphasis on
knowledge and continuous learning processes in the new forms of economic and social behavior that are
emerging. Effective learning and training is nowadays a necessary precondition to protect the citizens
against unemployment and social exclusions as the skills related to the use of information and
communication technologies become indispensable.
The parallel advances in information and communication technologies make it possible to develop new
kinds of educational and training services overcoming the constraints of traditional learning processes that
were monopolized by well established institutions. This is an opportunity that has already received much
attention and a lot of resources have already been invested in developing educational material as well as
establishing learning networks that bring together teachers and learners, trainers and trainees. Until recently
no significant efforts were taken to integrate all these systems and learning networks so that the learning
resources could be reused thus saving valuable resources. The proliferation of the Internet, the World Wide
Web and interoperability standards for digital learning resources create bright new application
opportunities for the future.
ADL's SCORM model (SCORM 2001) is probably the most important step towards the development of
interoperable eLearning systems by providing the means to overcome the incompatibility issues posed by
proprietary implementations through proper standard descriptions of learning content objects and related
material and functionalities. It is an effort supported by almost all the big players in the eLearning field that
gains thrifty acceptance in both commercial and academic environment. It provides reusability and
interoperability of learning resources by leveraging the advances in all fields related to eLearning through
proper integration and extension of various existing models. This strategy of integrating, instead of
substituting, existing popular learning resource specifications, amplifies its widespread acceptance. The
prototype implementation described in this paper is yet another exploration of the great potential of
SCORM and proves that it can effectively support the interoperability needs of an existing eLearning
platform with no compromises.
A major need shows up: How can we take advantage of the already existing learning resources in digital
form as well as the well established learning networks and at the same time take advantage of the
interoperability specifications that have emerged? The answer is not unique and usually depends on the
technical details of the existing infrastructure. A simplified classification of preexistent eLearning solutions
is to distinguish between coarse-grained file-based and fine-grained database-oriented approaches in terms
of learning resources. The first approach is more straightforward and monolithic, unable to provide
flexibility and personalized learning experiences. The second, more sophisticated and flexible, represents
also a greater effort in the design and implementation of the corresponding system functionalities as well as
an undoubtedly heavier authoring of learning resources. Having this in mind it is evident that eLearning
environments of the second kind are more interesting to investigate with respect to their interoperability
enhancements. Their complex structure makes it more difficult to support an interoperability specification
and their existing functionalities and learning resources represent a bigger investment that should be
preserved.
Talking in general, a fine-grained database-oriented eLearning environment could be conceived as three
layered system: At the lowest level, there is a (usually relational) database management system. At the
middle there is the specific database of the system along with a number of database transactions used to
store and retrieve data including software components for the creation of dynamic HTML pages in case of
web-based solutions. At the upmost level there are various user-centered applications providing functions
for browsing, authoring, user communication etc.
An effective strategy for supporting interoperability in such an environment should make the less possible
modification to the preexistent system. The proposed solution operates at the middle tier (the database of
the eLearning system and the dynamically created HTML pages) and it defines a mapping between the
logical model of the system with the content model of the SCORM. This is accomplished through a data
binder that provides a bridge between the SCORM model and the underlying database schema. An export
procedure picks up all the necessary data from the underlying database and then creates a manifest file
describing the organization and the contents of the content package that is being created. The raw material
of the package is collected piece by piece using an HTML parser that handles the (dynamic or static) pages
of the learning resources.
The rest of the paper presents first (Section 2) the generic architecture assumed in our work. Moreover, it
describes how the logical model of an elearning system can be mapped to SCORM and presents the export
process of learning resources to the SCORM format. Section 3 presents the prototype implementation on
top of a preexistent web-based eLearning platform and Section 4 presents some related work. Finally,
Section 5 concludes and presents the directions for future research and development activities.

2. SHARABLE CONTENT REFERENCE MODEL (SCORM)

SCORM (SCORM 2001) is probably the most important development currently occurring in the area of e-
learning standards and specifications. Simply stated, SCORM is a set of specifications for developing,
packaging and delivering high quality education and training materials whenever and wherever they are
needed. The use of SCORM enables reusability, accessibility and durability of the learning material in
technology changes, and interoperability between different e-learning platforms.
SCORM in its 1.2 version consists of two parts: The SCORM Content Aggregation Model and the SCORM
Run Time Environment. The SCORM Content Aggregation Model further defines a common way by
which learning content can be interoperable, interchangeable, reusable and accessible. The SCORM
Content Aggregation Model defines how learning content is identified, described, aggregated into a
“course” and moved between systems. Specifically, the learning resources comprising a learning
experience get packaged into a zip file (SCORM package of Package Interchange File (PIF)). This file
contains not only the course files, it also contains an XML file, referred to as the manifest file, describing
the course contents and content sequencing.
The SCORM Run-Time Environment defines the means by which SCORM learning content is made
interoperable between LMSs. The SCORM Run-Time Environment provides details on how LMSs should
launch resources/content, communicate with the content (using a predefined language and vocabulary) and
exchange predefined data elements during execution.

3. A GENERIC ARCHITECTURE
The proposed architecture supports the export of the educational content of an e-learning platform to
SCORM 1.2 Content Packages (IMS CP 2001), given that the content is stored in a relational database.
SCORM contains a rich dictionary of metadata terms that can be used for describing educational content.
This architecture supports also the transformation of the metadata contained in the relational database to the
format described in the IMS Learning Resource Meta-data Information Model (IMS LRM 2001). These
metadata, used to describe the learning resources, will be also included in the generated SCORM Content
Packages. The SCORM model has been selected, since –because it is widely adopted and implemented– it
will be undoubtedly the future eLearning standard. However, this architecture, with appropriate
modifications, can support other standards, if they use XML Schema for the description of the educational
content. In our case, the outcome of the export process will be –in terms of SCORM– the Package
Interchange File (PIF), a zip file that contains not only the physical files that compose an educational
experience, but also an XML file, referred to as the imsmanifest file, describing the course contents and the
content sequencing.
The architecture (Figure 1) follows a multi-tier approach with three distinct tiers:

Figure 1: A generic architecture for extending an eLearning platform to support SCORM

1. The existing e-learning platform.

2. The database tier, which is a relational database management system along with the relational
database, where the educational material and content metadata are stored.
3. The middleware tier, which realizes the transformation from database objects to the SCORM
objects. The middleware consists of two parts:
 a. The XML-DB middleware, whose role is the mapping between the relational database
schema and the two SCORM XML Schemas, one for the content packaging (IMS CP
2001) and the other for the description of learning resources with metadata (IMS
LRM 2001), and
b. The Package Management middleware, which is responsible for the transformation
process.
The existing e-learning platform can be any platform that is implemented on top of a relational database to
store all the information regarding the educational content. Both static and dynamic pages are supported in
this architecture.
The relational database is responsible for the storage and retrieval of information regarding the educational
content.
The transformation process is implemented through the middleware. For this to be possible, a mapping
between the RDB Schema and the two SCORM XML Schemas (based on IMS Content Packaging
Specification v1.1.2 and IMS Learning Resource Meta-data Specification v1.2) is needed, an XML
document management system on top of the relational database management system. This is the role of the
XML-DB middleware in this architecture. It is a set of software components responsible for the
manipulation of SCORM XML documents. In this case the data management software does not rely on
XML document modeling solutions (like DOM) but use another approach, called data binding (Bourret,
2001). Data binding offers a much simpler approach to working with XML and supports effective
separation between document structure and data modeling.
The term “XML data binding” refers to the binding of certain XML documents to objects. These objects
use a schema (classes) designed especially for the data in those documents. This allows applications to
manipulate data that has been serialized as XML in a way that is more natural than using the DOM. For
example, consider the following part of an XML document:

This could be bound to the Book, BasicDescription, Title and Author classes, so that when data is
transferred from the XML document, the result is a tree of objects:

To develop the XML-DB middleware we have used the help of a data binder to bind the SCORM XML
Schema to objects. There are numerous XML data binding products (Bourret, 2001) capable of transferring
data between XML documents and java objects. They are classified in two broad classes, namely design-
time binders (require configuration based on a DTD or an XML Schema before they can be used) and run-
time binders (used directly in code to serialize and de-serialize objects as XML elements). Design-time
binders are usually more flexible in the mappings they can support.
The architecture in Figure 1 assumes a design-time binder (Bourret, 2001). Thus, a configuration process
(illustrated with dashed arrows) is necessary to create the classes corresponding to XML elements present
in the document class to be handled. To be more specific, the two SCORM XML Schemas about the
Content Packaging and the description of learning resources with metadata (based on IMS Content
Packaging Specification v1.1.2 and IMS Learning Resource Meta-data Specification v1.2 respectively),
should be used, as depicted, to create two set of classes, each one (class) corresponding to a particular
element type. The class hierarchy that has been generated includes also marshal methods. Marshal methods
perform the serialization of the object tree to XML document and the whole process is called
“marshalling”.
The classes created during the data binding configuration process are extended with DBRetrieve methods.
DBRetrieve methods retrieve data from the database with the objective to build object trees that could be
used thereafter to create SCORM XML documents. DBRetrieve methods rely on both the class hierarchy
created by the data binding configuration process and the structure (relational schema) of the underlying
database. The communication with the relational database management system relies on the use of standard
interfaces like JDBC (Relational DBMS Access Interface in Figure 1). Standard SQL statements are used
to store/retrieve data from the underlying relational database.
A SAX-based parser and corresponding validator can be used to ensure that the generated XML documents
conform to the two SCORM XML schemas. The parsing process is closely related to the marshal methods.
An assumption is made here that the code generated by the data binder handles all the necessary
interactions with the parser.
The Package Management middleware does all the remaining work needed for the transformation. In
practice, it manages the export process. Hence, it is responsible for finding all learning resources that
compose an educational experience, and store them in form of SCO’s and Assets inside the SCORM
Content Package along with the metadata in appropriate format that describe them. The functionality of this
part will be analytically explained later in this paper, where the whole export process is described.

3.1 Mapping the logical model of an eLearning system to SCORM

Before proceeding to the implementation of the transformation system, a mapping between the eLearning
platform’s elements and SCORM elements must be found. With simple words, one must firstly define
which are the lowest level components or the minimum granularities of educational content that might be
re-used in other learning experiences. Each of these components should provide useful learning content by
itself and must be independent of learning context, to be reusable. If we add one more characteristic to the
above, which is that these components must be designed to be launched and tracked by a SCORM-
compliant LMS, we reach the SCORM definition for Sharable Content Objects (SCOs). All learning
resources from which one SCO consists of, are the SCO’s Assets. For example, some times an individual
web page might be considered as a SCO, but usually more than one page is required to convey meaningful
learning content. Once we have decided which are the SCOs and added some code for the SCO – LMS
communication, we must also decide which will be their organization or organizations and sequencing
inside a learning experience that will be packaged in SCORM format. The existing presentation order can
be followed for the organization and sequencing of the learning resources inside the generated SCORM
package.
The above points must have been considered, since they determine how the DBRetrieve methods will be
constructed. In addition, the mapping between the DB elements and SCORM regarding the metadata
should be considered. The appropriate values should be retrieved from the database to “fill in” the
corresponding elements and attributes in SCORM XML Schemas.

3.2 The export process

As we already mentioned, the export procedure includes the transformation of the educational content that
composes a learning experience and metadata describing it, which are stored in the RDB, to the SCORM
1.2 format. The outcome of this process is the creation of a SCORM 1.2 Content Package, which contains
all learning resources (physical files – SCOs and Assets) composing the learning experience, and the
imsmanifest.xml file, which describes these learning resources and their sequencing. The metadata that
describe the learning resources can reside as inline data into the imsmanifest file, or in separate XML
documents, one for each learning resource. In the last case, pointers are included in the imsmanifest file to
the corresponding XML documents. Both cases can be supported.
Once the export procedure has been started, the DBRetrieve methods are sequentially executed and they
build firstly the “manifest” and “organizations” objects. The number and the structure of the “organization”
objects inside the “organizations” object depend on the way the DBRetrieve methods have been written.
In an eLearning platform, static or dynamic pages can be used to present the educational content to the
learner. In the first case, the same static pages can be stored inside the SCORM Content Package to present
the learning experience to the learner. In the second case, dynamic pages could not be saved inside the
SCORM Content Package and used as they are, because if so the SCOs would not be context-aware.
Hence, a mechanism should exist to store in Content Package the result of the execution of the dynamic
pages in the server, which has the form of static pages. This mechanism is supported in this architecture (in
Package Management Middleware) (Figure 1).
Another point is that a web page (either it is considered as a SCO or an Asset) may contain images, links to
other files or even references to files that may not be apparent in page’s interface, but are referenced or
included in the HTML code (e.g. Stylesheets, JavaScript files). These files should also be stored in
appropriate places in the Content Package, so that the pages could be presented correctly and transitions to
other pages or files could be performed. For this reason, our architecture includes (in Package Management
middleware) a mechanism that finds all pages’ dependencies, downloads all referenced files and stores
them in appropriate places inside the Content Package. An HTML Parser and an XSLT Processor are
needed and the whole procedure is illustrated in Figures 2 and 3.

Figure 2: Detection of all pages’ dependencies

Figure 3: Serialization of the DOM Tree to HTML and storage of the generated HTML page in the package

In the DOM Tree that is generated from the HTML page code, all references to downloaded files are
converted into relative paths. Moreover, some new elements can be automatically added to the DOM Tree,
in order to add code for the SCO-to-LMS communication.
Instead of downloading and saving in the Content Package all the physical files needed, pointers can be
placed inside the manifest XML document that point to the files’ URLs. In this case, these files are
“external resources”.
After finding and storing in the Content Package all the physical files that are needed in a learning
experience or pointers to the corresponding URLs, in case they are external resources, the “resources”
element is created, which contains and describes all resources that compose the learning experience. These
are SCOs or Assets. The information about the “resource” elements is known from the DB and the previous
procedure.
Finally, the manifest instance gets marshalled (output as XML) to a file (imsmanifest.xml), inside the
content package and the contents of package get zipped to a file (Package Interchange File), that can be
imported to any SCORM compliant LMS.
For example, in Figure 4, a SCORM package, which has been generated from our implementation
described in the next section, has been successfully imported in the ADL’s Sample Run-Time Environment
v1.2.1. Moreover, our generated packages have been tested with SCORM Version 1.2 Conformance Test
Suite Version 1.2.3, which is provided by ADL, to ensure that they are 100% SCORM conformant.
 Figure 4: Import of a generated package to the ADL’s Sample Run-Time Environment v1.2.1

4. PROTOTYPE IMPLEMENTATION
The proposed architecture has been successfully implemented on top of an existing e-learning web-based
system, the Distant Learning Center (DLC) (Stellakis, 2000). DLC is a system that provides distance
training through the use of the World Wide Web (Web-based training) and has been developed in the
Laboratory of Distributed Multimedia Information Systems and Applications of the Technical University of
Crete (TUC/MUSIC) for the needs of the European ARCHIMED project (Advanced Multi-media-System
Architectures and Applications for Educational Telematics, INCO-COPERNICUS PL96106). DLC is
based on Internet and on a web server for the information delivery to its users’ browsers. It is very flexible,
since all the information is stored in a relational database. The communication with the relational database
management system relies on the use of an ODBC interface. HTML, DHTML and JavaScript have been
used for the content presentation to the browsers and Active Server Pages (ASP) for dynamic pages’
development.
There are two user categories in DLC: The Learners, which attend the courses and participate to the
learning activities provided by the system, and the Instructors, which develop, organize and supervise these
courses.
Learners can attend a course and see the courseware in an environment where they can browse content, see
examples, and see related videos. They can also add annotations on the courseware, which can be visible
only to them or to all users. Moreover, they can add annotations on annotations. The collaboration and
communication between learners take place either through these annotations, or through their participation
in discussion forums, places where learners can interchange opinions and ask questions by messages visible
to all users. Learners can, also, be organized into groups, having this way the capability for a better
collaboration. Finally, learners can take exams for each course. Their evaluation is made automatically
from the system or after the instructor’s mediation depending on the type of the exercise or the exam.
Instructors add courseware, examples, exercises and exams. They evaluate the learners’ answers in these
exercises, and, also, construct and supervise the discussion forums of their courses. Moreover, they
supervise the whole course, attending the public annotations of the learners or adding their own
annotations.
The personal communication between the members of the system (learners and instructors) takes place
through private messages, which are text messages that apply to one recipient.
The courseware in DLC, can be organized either as a book (with chapters, sections, and subsections) or as a
lecture (with a series of slides referring possibly to the components of a book). The same book organization
can support numerous lecture organizations. Moreover the DLC has a digital library of documents or files
with content related to the courses.
Private or public annotations can be added to text objects, audio pages, video pages and examples. The role
of an annotation is to explain or comment the object with which it is related.
For the learners’ assessment, each course has a collection of test objects (Questions and Exams). The
questions are related with the course’s sections. They are of four kinds: Unique answer, True/false, multiple
choice and Free answer questions.

4.1 Mapping the logical model of DLC to SCORM

In DLC, the chapters, sections, subsections and the presentations can be considered as candidates for
becoming SCOs. Searching for the minimum content granularity that can be reused in another learning
experience, one can easily see that the best selection is slide objects. If instead another class of objects (e.g.
chapters) was selected, it would not be functional, since we could not reuse its lower components. It should
be nevertheless stressed that one slide does not always have independent educational content. It could be
part of a cohesive unit of slides that cover one educational issue. Although a slide is not always
independent, we would like to be able to reuse this slide in other courses composition
Once we have defined the learning resources, we must decide which will be their organization or
organizations. The two organizations of DLC (book and lecture) will be followed to organize the
educational content inside a SCORM package. In figures 5 and 6 the mapping between DLC and SCORM
elements in both cases is illustrated.

Figure 5: Mapping for book organization

 Figure 6: Mapping for lecture organization

Then, we proceed to the construction of DBRetrieve methods and we use the above architecture for the
transformation process. The export and import of all kinds of Annotations is also supported. Annotations
are considered as Assets. In case of a text annotation, a web page is stored in content package. If the
annotation is an audio or video file, it can be placed also in content package, or a pointer to its URL can be
used (external resource).

5. RELATED WORK
In general, describing a course through SCORM needs reconstruction of the course almost from the
beginning. This process can be performed either at low-level, by finding a mapping of existing resources to
SCORM content model components, writing appropriate XML documents and adding physical files to
construct SCORM Content Packages, or at a higher level, using existing tools to design SCORM
conformant courseware, as Aspen ToolBook Instructor. In the paper of Qu Ch. (Qu and Nejdl, 2002), a
SCORM conformant courseware has been implemented, by re-designing all learning resources contained in
an existing Java course according to the SCORM Content Model and also reconstructing the course
structure according to the SCORM Content Packaging. Considering a big collection of courses that have
been developed in one learning platform, this manual redesigning of each course from the beginning would
be a time and cost ineffective procedure. Instead of re-designing all courses that have been developed in a
specific learning platform, the proposed architecture can be used to extend the platform’s functionality
performing this conversion process automatically.

6. CONCLUSIONS
To maintain and extend the value of preexistent investments in eLearning systems and preserve digital
learning material there is a need to support interoperability specifications and emerging standards that
allow learning resource reusability and integration of operational eLearning environments. The
methodology and architectural considerations reported in this paper address this need of extending
preexistent eLearning environments in order to support the most promising and successful interoperability
specifications, which constitute the SCORM suite.
The baseline of this work is a generic reference architecture that assumes a sophisticated preexistent
eLearning system based on a relational database management system. We propose effective mechanisms
for transforming the contents of this system in the standard form proposed by the SCORM specifications.
We also address the major implications and implementation issues and describe a successful
implementation on top of an existing e-learning web-based system, the Distant Learning Center (DLC).
It is evident that the feasibility of the proposed approach constitutes a successful evaluation of the SCORM
specifications and enables its faster proliferation as it puts forward a safe transition pattern for existing
systems and learning resources repositories.
DLC, empowered with SCORM export functionality is used in two European vocational training networks:
The first network addresses the needs of workers in the tourism and cultural sector (ADONIS) and the
second focuses on the mass media industry (KNOSOS).
At this time, our system does not support the export or import of questions or exams. The main reason is
that SCORM 1.2 does not contain a schema or guidance for the description of assessments. Assessments
constitute a special kind of educational content and need special treatment. IMS Global Learning
Consortium has developed the Question and Test Interoperability (QTI) specification, which defines an
XML format for the coding of questions and exams. This specification makes able the transfer of such
kinds of objects between different LMSs and is very possible in the future to be a part of SCORM.
One major direction of future research is the investigation of interoperability between SCORM and
audiovisual standards. A first step towards this direction has been reported in (Frantzi et al, 2004). That
work investigated the interoperability issues between SCORM and the TV-Anytime Metadata Model in
order to provide effective specifications and system architectures that will transform digital TV into a
learning medium thus opening new opportunities to the broad public. We believe that such integration will
provide new delivery channels for learning services including ubiquitous computing devices allowing a
transition from current situated learning paradigm to what we can call "ubiquitous learning". We are
currently working on exploiting interoperability issues between SCORM and MPEG7.

REFERENCES
1. Arapi P., Moumoutzis N., Christodoulakis S. (2003). Supporting Interoperability in an Existing e-Learning
Platform using SCORM, Proceedings of the 3rd IEEE International Conference on Advanced Learning
Technologies (ICALT 2003), Athens, Greece
2. Bourret R. (2001). XML Data Binding Resources, URL:
http://www.rpbourret.com/xml/XMLDataBinding.htm
3. Frantzi M., Moumoutzis N., Christodoulakis S. (2004). A Methodology for the Integration of SCORM with
TV-Anytime for Achieving Interoperable Digital TV and e-Learning Applications, Proceedings of the 4th
IEEE International Conference on Advanced Learning Technologies (ICALT 2004), Finland
4. IMS CP (2001). IMS Content Packaging Specification v1.1.2,
http://www.imsproject.org/content/packaging/index.html
5. IMS LRM (2001). IMS Learning Resource Metadata Specification v1.2.1,
http://www.imsproject.org/metadata/index.html
6. Qu Ch., Nejdl W. (2002). Towards Interoperability and Reusability of Learning Resource: a SCORM-
conformant Courseware for Computer Science Education, Proceedings of the 2nd IEEE International
Conference on Advanced Learning Technologies (ICALT 2002), Kazan, Tatarstan, Russia
7. SCORM (2001). Sharable Content Object Reference Model, Version 1.2, U.S. ADL
8. Stellakis, D. (2000). Design and implementation of a distance learning environment on the web, Diploma
thesis (in greek), Electronic and Computer Engineering Department, Technical University of Crete, Chania