(Departement Computerwetenschappen
Katholieke Universiteit Leuven, Belgium
Erik.Duval@cs.kuleuven.ac.be)



  In order to be able to (re-)use digital content, interested users must be able to identify
and locate relevant documents. This requires descriptive data, nowadays generally referred to
as
. Technical  for a scaleable deployment on a global scale are required if
we want to achieve a critical mass of resources. In this paper, we present the current status of
ongoing work in this area, with a particular emphasis on the IEEE LTSC Learning Object
Metadata standard [IEEE, 2001] and related developments in the context of the ISSS Learning
Technologies Workshop [ISSS, 2001].

G46 
metadata, learning technology standardization
  H.3 - Information Storage and Retrieval



 
! " 

Metadata is often defined as:




.

A somewhat more informative  is [IEEE, 2001]:











.

Thus, metadata are basically descriptive data. As such, metadata are at the heart
of more general developments in the area of digital libraries [Fox et al., 2001]. Basic
metadata elements indicate the title, author, year of publication and similar simple
bibliographic data. Richer metadata structures also cover technical features, copyright
properties, annotations and so on.
The 
 of metadata is ‘to facilitate search, evaluation, acquisition, and use’
of resources [IEEE, 2001]. Moreover, in the case of educational resources, the
purpose is also ‘to facilitate the sharing and exchange of learning objects, by enabling
the development of catalogs and inventories while taking into account the diversity of
cultural and lingual contexts in which the learning objects and their metadata will be
exploited’ [IEEE, 2001].
In the documents on metadata and the ‘semantic web’ from the web consortium,
metadata is often used for descriptive data that can be processed by machines
[Berners-Lee et al., 2001]. This is a more restricted interpretation than the one we
adopt here. By explicitly including descriptive data that need to be interpreted by
humans, we want to recognize the importance and relevance of such metadata.
Journal of Universal Computer Science, vol. 7, no. 7 (2001), 591-601
submitted: 1/5/01, accepted: 1/6/01, appeared: 28/7/01  Springer Pub. Co.

We will explicitly  focus on metadata for geo-spatial applications, library
applications (Z39.50, MARC variants) and continuous media specific standards (such
as MPEG-7), though most of what we will present here is applicable to those more
specific contexts as well. This paper will also not deal with the 



specification, that defines 15 elements for cross-domain search. The Dublin Core
specification has recently been submitted for approval as an American national
standard, referred to as Z39.85. There is a Memorandum of Understanding between
the IEEE LTSC LOM group (see below) and the Dublin Core initiative, with the
intent to investigate common mechanisms for interoperability between the two
metadata schemes, potentially based on an interoperable approach that builds on the
RDF framework of the World Wide Web Consortium. More details on how Dublin
Core relates to the IEEE LTSC LOM specification can be found in [Duval, 2001].


#
 
! $ 

#% ! 

Generally speaking, technical standards are important because they make it possible
to develop interoperable tools and services [Paepcke et al., 1998]. In this context, my
favourite definition of  is [Rust & Biede, 2000]:





























. There are a number of noteworthy aspects in this definition:

• The central notion is that of crossing boundaries of : this may involve
straightforward technical boundaries (like when metadata are served from a
server by a particular vendor to a client from another origin), but also more
subtle boundaries, such as linguistic ones (like when metadata are to be
translated), social ones (like when metadata intended for teachers need to be
transformed into metadata for learners), or, more generally, cultural ones
(like when metadata refer to national or regional educational contexts, such
as ‘bac+2’ in France). It is clear that the technical boundaries are the easier
ones to cross.

• The definition above mentions ‘as highly automated 
’. Obviously,
it is to be preferred that the process of crossing context boundaries is fully
automated, as in the case when documents are translated from one format
(like LaTeX or Microsoft Word) into another one (like HTML or Adobe
Portable Document Format). However, the definition makes it clear that this
process is not always fully automatic, as for instance in the case when
examples in a document need to be replaced by examples from another
application domain. (For instance, a document on the concept of
‘calibration’, originally developed for the automobile manufacturing industry
may be reused in the context of medical measuring equipment.)
592 Duval E.: Metadata Standards: What, Who and Why

Thus, the notion of interoperability is not a binary one, where systems would be
either interoperable or not. Rather, there is a higher or lower degree of
interoperability. This is well illustrated by the standards on paper size: in Europe, the
‘DIN A4’ standard, from German origins, prevails, whereas, in the United States, the
‘U.S. letter’ standard is more widespread. These standards have become so widely
accepted that they are almost ‘invisible’: we assume that papers fit in binders, that
binders fit in closets, that paper trays of printers have the correct size, etc. Moreover,
we all take it for granted that we can buy the hardware involved from different
companies and make the different components work together without any
transformation. Nevertheless, many of us have the experience of printing that goes
less than perfect when we download and print documents that have been formatted for
the ‘other’ standard size.
The above illustrates the main advantage of interoperability: it prevents end users
from being locked into proprietary systems. The World-Wide Web is a perfect
example of how standards (in this case: URL, HTTP and HTML [Berners-Lee, &
Fischetti, 1999] can be the basis of open, interoperable systems, that allow end users a
choice of client and server systems alike. The Web also illustrates that interoperability
is not always absolute: because of the diverging additions to the official W3C
standards that Netscape and Microsoft support, some features may only be available
on one platform, or the developers may be required to develop those features in non-
standard ways, separately for each supported platform.


#%#&
' $ 

It is useful to distinguish different levels of interoperability [see table 1]. At the most
machine oriented level, there is network protocol interoperability, where the relevant
standards include TCP/IP and HTTP. The HTTP standard for instance enables a Web
browser and server to exchange messages, even when these software components
were developed by different vendors, operating under different operating systems, on
different kinds of hardware, etc.
Secondly, there is the level where data gets bound to a particular representation
format or data binding. A typical web example is the representation of a document in
HTML. For metadata, the most popular bindings nowadays are XML, or, more
specifically, RDF [W3C, 2001].

1 Protocol TCP/IP, HTTP
2 Data binding HTML, XML, RDF
3 Metadata scheme LOM, Dublin Core
4 Semantic Ontologies, classifications, vocabularies, taxonomies

( : Layers of interoperability

The level that we will focus on in the remainder of this paper is that of the
conceptual data model or metadata scheme, that specifies the data elements of which
a metadata instance is composed. Metadata instances based on a common metadata
schema have a high degree of ‘semantic interoperability’ [Forte et al., 1999]. The
593Duval E.: Metadata Standards: What, Who and Why

binding of metadata schemes in level 2 representations is typically defined in a
binding specific way. As an example, an XML DTD has been developed for the IEEE
LTSC LOM specification, in order to define an XML binding of LOM. Similarly,
alternative mechanisms can be used to bind to the same representation format (for
instance: XML Schema) or alternative representations (for instance: RDF or SQL
schemas). In [Section 3], we deal with standards for layer 3 interoperability.
Finally, ontologies, classifications, vocabularies and taxonomies attempt to define
common semantics. In most cases, these conceptual structures are restricted to a
particular domain. Typically, they define the relevant concepts in that domain, and
their interrelationships. The intent is to enable consistent interpretation of statements
that make use of these concepts. An example of this layer of interoperability is the
reference ‘Category H.3 - Information Storage and Retrieval’ in the header of this
paper: it refers to the ACM Computing Reviews classification widely adopted in the
domain of computer science. Common classification structures are the basis of
consistent descriptions that support systematic access for indexers and end users.
More sophisticated such approaches are based on knowledge engineering
technologies [Hill et al., 2000].


)

* '
 


)% 
* 

Three ‘official’  standardization organisations are active in the field of
educational technologies in general, which includes the more specific field of learning
object metadata. These organizations are:

1. The 


!"



(LTSC) was set
up in 1996. Its purpose is to standardize the ‘smallest, useful, doable
specification that has technically feasibility, commercial viability, and
widespread adoption’. Besides working groups on for instance ‘Computer
Managed Instruction’, ‘Simple Identifiers’ and others, there is a group that
focuses on ‘Learning Object Metadata’ (LOM) [IEEE, 2001].
2. The Centre Européen de Normalisation organizes a #




since 1999, under the umbrella of the so-called ‘Information
Society Standardization System’ (ISSS). The main purpose of this workshop is
to ‘promote the development and adoption of appropriate standards, taking into
account the diversity of cultural backgrounds and languages that exists within
Europe’. After an initial requirements analysis [ISSS, 2000], work has now
started on LOM related work (see below), copyright, quality issues,
educational modelling languages, etc. [ISSS, 2001].
3. ISO and IEC have set up a Joint Technical Committee (JTC1) that, since 1999,
has a 





. At this moment, this more
formal body is initialising its operations. In the domain of metadata, it has
invited the IEEE LTSC to submit its LOM standard as soon as LTSC deems
appropriate.

594 Duval E.: Metadata Standards: What, Who and Why

Besides the formal standardization bodies, there are numerous  that
carry out technical work in the field of educational technologies. Once this work leads
to mature specifications, those specifications can be submitted to the accredited
standardization organizations. Conversely, consortia often represent communities of
practice that adopt standards as they are developed by accredited organizations.
The consortia with a more direct standardization impact on metadata include:

1. The $%$& Foundation regroups academic and industrial members
[ARIADNE, 2001]. At the core of its infrastructure is the so-called Knowledge
Pool System, a distributed repository of pedagogical documents and their
associated metadata [Duval et al., 2001]. An integrated Web-Based Learning
Environment supports the development of courses that reuse resources from
the Knowledge Pool System.
2. The '! consortium regroups vendors of Learning Management Systems,
authoring tools, and related products [IMS, 2001]. IMS does not develop
implementations, but focuses on the development of specifications that can
then be submitted to the standardization bodies mentioned above. Work is
currently ongoing in the area of content packaging, question and test
interoperability, etc.
3. $ was originally a U.S. Army initiative for interoperability developments in
the area of learning technologies, but it has substantially increased its scope
and relevance. One of its major milestones is the Sharable Content Object
Reference Model [ADL, 2001]. The major aim of the SCORM model is to
define an overall specification for interoperability between components of a
digital learning infrastructure, based on the IEEE LTSC LOM and CMI
standards.

The IEEE LTSC LOM standard is based on early work by the ARIADNE and
IMS consortia, which led to a joint submission of a base document. Since then,
ARIADNE, IMS and ADL have all contributed to further development of the LOM
specification within the IEEE working group. At the time of writing, the LOM
standard is in ballot.
ARIADNE, IMS and ADL are now developing their own so-called ‘application
profiles’ of the LOM standard: these are specifications that adapt the standard to the
specific needs of their communities. In practice, this can for instance involve a
mandatory status for some data elements, or more restricted vocabularies than those
contained in the IEEE specification (see below).


)%#
& *+


)%#% & *+


In the context of the IEEE LTSC LOM, the term "Learning Object" should be
understood in its most general sense. The definition in the standard is:

(






















)*
595Duval E.: Metadata Standards: What, Who and Why

Thus, learning objects can be of any size, type, etc. In principle, they need not be
digital, and can include people, rooms, equipment, etc. This concept of a generalized
learning object is in contrast to that of an object as a discrete item or piece of content,
often within a hierarchical content model that progresses from the level of raw media,
up through content objects, learning objects and then lessons, courses, curricula, etc.
Moreover, a learning object need not be restricted to a static object or piece of
content: it can also be a momentary collection or assembly of content, for instance
adapted to the specific needs of a particular learner in a given situation and time.

)%#%#"

,

The IEEE LTSC LOM standard defines a so-called 

. This is basically a
collection of data elements that can be used to describe a learning object. The LOM
scheme regroups data elements in nine categories:
1. The + category groups information that describes the learning object as
a whole. This category includes elements like identifier, title, language,
keywords, etc.
2. The  category groups the features related to the history and current
state of the learning object. It also describes the individuals or organizations
that have affected the learning object during its evolution. Data elements in
this category include the version, status, and contributors (authors, publishers,
etc.).
3. The ',
 category groups information about the metadata, rather
than about the learning object that they describe. This includes an identifier for
the metadata instance, contributors to the metadata, the language used in the
metadata, etc.
4. The  category groups the technical requirements and characteristics
of the learning object. This category describes for instance the MIME type of
the learning object, its size, location, required soft- and hardware, etc.
5. The 
 category groups the educational and pedagogic characteristics
of the learning object. It indicates the interactivity type (active, expositive,
etc.), learning resource type (exercise, simulation, questionnaire, etc.),
interactivity level, semantic density, educational context (primary education,
higher education, vocational training, etc.), typical age range, etc.
6. The % category groups the intellectual property rights and conditions of
use for the learning object. For this category, LOM adopted a fairly simple
approach, indicating whether or not any cost is involved, and whether
copyright and other restrictions apply. The idea is to refer to other standards
for more complex modelling of rights management metadata [INDECS, 2001].
7. The % category regroups features that define the relationship between
this learning object and other ones, with an indication of the type of the
relationship (‘based on’, ‘part of’, etc.).
8. The $ category provides comments on the use of the learning object
and information on when and by whom the comments were created.
9. The  category describes where the learning object can be
classified within a particular classification system. As any classification can be
referenced, this category provides for a simple extension mechanism.
596 Duval E.: Metadata Standards: What, Who and Why

3.2.3 Data Elements

For each data element, the base scheme defines:
• 
- the name by which the data element is referenced;
• - the definition of the data element;
• "- the number of values allowed;
• - whether the order of the values is significant (only applicable for data
elements with multiple values);
• .

- the set of allowed values for the data element - typically in the
form of a vocabulary (see below) or a reference to another standard (such as
vCard, ISO8601 for the representation of dates, etc.);
• 
- a set of distinct values;
• 
: an illustrative example.

Some data elements contain 
,
. Data elements with sub-elements do
not have values directly, but indirectly, through their sub-elements. As an example,
the element that indicates the learning object that the described object is related to
(Relation.Resource) has a value indirectly only, through one or more of its
subelements (Relation.Resource.Identifier, Relation.Resource.Description or
Relation.Resource.CatalogEntry).
/
 are recommended lists of appropriate values, that define the value
space of a data element. Other values, not present in the list, may be used as well.
However, metadata that rely on the recommended values will have the highest degree
of semantic interoperability, i.e. the likelihood that such metadata will be understood
by other end users is highest. As an illustration, the data element
Educational.LearningResourceType may have a value from the LOM vocabulary,
such as for instance "Questionnaire". This option is preferred if the values in the
vocabulary can adequately express the intended meaning. If the indexer wants to
assign a value that is not part of the list given in the LOM document for that data
element, then the indexer may designate the value as, for instance,
("http://www.vocabularies.org/ LearningResourceType", "MotivatingExample"). This
option provides more flexibility to the indexer of learning objects, at the expense of
semantic interoperability. User defined values will not be used consistently
throughout the larger community. In the example above, a URI was used to indicate
the source of the vocabulary. This approach is certainly good practice, but using a
URI is not a requirement.
For each of the data elements, the specification includes the 
 from which
it derives its values, such as Date, Character string, etc. Of particular interest is the
notion of ‘LangString’, used to represent a phrase in a human language. A value of
this type can consist of multiple (Language, String) tuples where Language indicates
the human language (according to the ISO639 standard) and String holds the actual
character string (according to ISI/IEC10646-1). An example of this concept, as
represented in an XML binding could be:
597Duval E.: Metadata Standards: What, Who and Why

<title>
<langstring>
<string xml:lang="en">Draft Standard for Learning
Object Metadata</string>
<string xml:lang="nl">Voorstel van Standaard voor
Metadata van Leerobjecten</string>
</langstring>
</title>

In this case, two titles are defined for the learning object: one in English, and one
in Dutch.
A LOM metadata instance may contain 
data elements. Such elements
cannot replace data elements in the LOM structure.


)%) $'' 
&*


The IEEE LTSC Learning Object Metadata schema is explicitly recognized by the
ISSS Learning Technologies Workshop as the commonly accepted global standards
solution for describing learning objects through metadata. The Learning Technologies
Workshop is complementing this global activity with a number of projects that
address Europe’s specific requirements [ISSS 2001]:

• A first project will ensure that the IEEE LTSC LOM, as the globally
accepted solution, is capable of addressing specific European cultural
requirements (such as multilinguality). The outcome of this project may be a
proposed addendum for LOM.
• A second project is investigating standardization actions to permit the
identification of alternative versions of resources, in different languages, as
well as the origin of the translation, all within a LOM context. The outcome
may be an application profile of LOM to deal with this specific requirement.
• A third project will ensure that LOM is localized and translated in the
languages of the EU and EFTA countries. Translations of earlier versions of
LOM already are available. These will be replaced in due time by updated
and widely accepted revised versions.
• On the semantic level of interoperability, the workshop will collect and
organize a registry of taxonomies and repositories relevant to a European
learning society, via an on-line repository. The registry will indicate the
applicability of taxonomies and vocabularies, their interrelationships, as well
as mappings and translations between different structures. This will benefit
interoperability between European learning technology systems and services
as metadata implementations will be able to rely on standardized taxonomies
and vocabularies. It is expected that many will be developed and
implemented at national level. Actions will focus on the identification of
existing taxonomies, their applicability and interrelationships. Where
possible, mappings or translations will be made between various taxonomies
and vocabularies used in multilingual and multicultural learning domains.
598 Duval E.: Metadata Standards: What, Who and Why

All in all, it looks likely that LOM will be widely adopted in Europe, as this
standard is well suited to deal with the multilingual European context. This is not
surprising, as much of the original research and development took place under the
European ARIADNE umbrella.


-*$


$ ,


We believe that the LOM standard provides a sound basis for educational metadata.
Even if we take this for granted, a series of important issues and problems require
further attention.
Awareness about the relevance of metadata for knowledge management in
general, and for educational purposes in particular, has increased sharply these last
years. In itself, this is obviously a positive evolution. However, it also raises issues
about 


 towards the community of end users and developers,
and about 


. Even though most of the relevant organisations
have a very open approach, where basically anyone can participate and contribute, the
technical nature of the work and the somewhat obscure formalisms involved
(‘acronym soup’) may make the field somewhat intimidating to newcomers or those
who just want to assess the impact on their own work. Moreover, an analysis is
required of how different communities adopt and support educational metadata for
their constituencies [Duval, 2001].
Authoring of data and metadata is (too) 





: automatic
generation of obvious metadata is useful and possible, but especially semantic
metadata will in most cases need to be provided through human intervention.
Metadata templates can help to make this process of indexation easier, especially
when similar documents need to be described regularly. Moreover, the development
of interesting educational resources, that  add value when compared with their
paper counterparts (books, slides, etc.) is extremely time consuming and quite
complex, the more so as it often requires a multidisciplinary team of context experts,
graphical designers, technical experts, pedagogical experts, etc.
We basically argue that standardized metadata help end users to identify and
locate relevant educational material. However, that doesn’t mean that, once such
material has been identified, no 



0,1
 remain: the user interface
or look-and-feel of the resource may need to be adapted to the overall context it will
fit in, there may be technical, organisational or legal reasons that prevent the material
from being made available to new users, the pedagogical style of the resource may not
be appropriate for the intended new context, etc. This issue raises the question of
adaptation of resources, either through human intervention (which requires
interoperable authoring environments) or automatically (which is beyond the current
abilities of so-called adaptive systems).
In the more general sense, there is the open question on 





", and in what order. This is to be considered in the context of political,
legal and other sensitivities. The question of appropriate priorities is even more
important when one realizes that the standardization process takes a long time:
between the first stable ARIADNE metadata specification and the LOM standard
ballot, five years were spent on consensus building, evaluations and testing!
599Duval E.: Metadata Standards: What, Who and Why

Finally, there is the issue of interoperability in a wider sense: as the goal is to
realize an infrastructure for interoperable tools and services, the question arises what
the appropriate building blocks or components for such an infrastructure are. Should
there be document and metadata servers? Should these be the same? Can their data
modelling and management requirements be met by traditional database technologies?
Should we rather opt for a peer-to-peer approach? How will management of
educational resources tie in with general knowledge management? What about
security? Etc. Etc.


.


In this paper, we have argued that standardized metadata are a prerequisite for large-
scale deployment and (re-)use of educational resources. The standardization process
in this area is maturing rapidly, with the first stable specification (IEEE LTSC LOM)
now under ballot. This seems to suggest that the first requirement for a worldwide
pool with a critical mass of reusable pedagogical documents can be met. This
situation creates exciting opportunities for further research and development in this
area (design for reuse, semantic interoperability, metadata and document authoring).


/' 


[ADL, 2001] http://www.adlnet.org/.
[ARIADNE, 2001] http://www.ariadne-eu.org/.
[Berners-Lee, & Fischetti, 1999] T. Berners-Lee & M. Fischetti. 2.

2-


3


4




2
2
2



.) Harper, 1999.
[Berners-Lee et al., 2001] T. Berners-Lee, J. Hendler and O. Lassila. 
!


2. Scientific American, May 2001
<http://www.scientificamerican.com/2001/0501issue/0501berners-lee.html>.
[Duval et al., 2001] E. Duval, E. Forte, K. Cardinaels, B. Verhoeven, R. Van Durm,
K. Hendrikx, M. Wentland Forte, N. Ebel, M. Macowicz, K. Warkentyne, F.
Haenni, 
$%$&
G46
6
!
-








, Communications of the ACM, May 2001.
[Duval, 2001] E. Duval. !"
'


-

!

%,
EdMedia2000, 26 June-1 July 2001, Tampere, Finland.
[Forte et al., 1999] E. Forte, F. Haenni, K. Warkentyne, E. Duval, K. Cardinaels, E.
Vervaet, K. Hendrikx, M. Wentland-Forte, and F. Simillion, !



6

'



$



%
ACM SIGMOD Record Vol. 28, no. 1, 20-25, March 1999.
[Fox et al., 2001] E. Fox & G. Marchionini. !




)
Communications of the ACM, May 2001.
[Hill et al., 2001] L. Hill and T. Koch (eds.), &#
G46
3

!
-






, Journal of Digital Information,
Volume 1, Issue 8 <http://jodi.ecs.soton.ac.uk/Articles/v01/i08/editorial>.
600 Duval E.: Metadata Standards: What, Who and Why

[IEEE, 2001] 
!


3
' draft 6.1, April 2001
<http://ltsc.ieee.org/wg12/index.html>.
[IMS, 2001] http://www.imsproject.org/
[INDECS, 2001] http://www.indecs.org/.
[ISSS, 2000] $
!"
2#
6



(




7


'


!8)
CEN Workshop
Agreement 14040, 14 July 2000.
[ISSS, 2001] <http://www.cenorm.be/isss/Workshop/lt/>.
[Paepcke et al., 1998] A. Paepcke, C. K. Chang, H. Garcia-Molina & T. Winograd.




2) Communications of the
ACM, Volume 41, No. 4, pp. 33-43, April 1998.

[Rust & Biede, 2000] G. Rust & M. Bide. 
9:



#)

6




) June 2000
<http://www.indecs.org/pdf/framework.pdf>.
[W3C, 2001] '

%

, April 2001
<http://www.w3.org/Metadata/>.
601Duval E.: Metadata Standards: What, Who and Why