Semantic learning object repositories


Jesús Soto Carrión
Elisa Garcia Gordo
Languages and Informatic Systems Department,
Pontifical University of Salamanca,
Madrid Campus, Spain
E-mail: {jesus.soto, elisa.garcia}@upsam.net
Salvador Sánchez-Alonso
Information Engineering Research Unit,
University of Alcala, Spain
E-mail: salvador.sanchez@uah.es


1. Introduction
A Semantic Learning Object Repository (SLOR) is a software system that stores educational resources and
their metadata (or only the metadata), and provides some type of searching interface to human or to other
software systems. Repositories provide access to the collections of educational resources, usually in electronic
format, although most of them do not store the educational resources themselves, but only their metadata.
Therefore, it is possible to find the same resource from different repositories. The main functionality of a
learning object repository is that of allowing for the search of educational resources. From this perspective
and, in a broad sense, the following two types of repositories exist:
1 Interactive searching interfaces, to be used by humans.
2 Querying interfaces, mostly aimed at being used by software agents (for example, through web
services).
Sometimes, the same repository includes a searching method that can be used to both usages. Nevertheless, it
must be noted that most usual general propose information retrieval mechanisms (Baeza-Yates and Ribeiro-
Neto, 1999) (like internet searching) must be complemented with metadata browsing mechanisms. The
simplest way to implement this is allowing metadata field searching. However, very often those interfaces are
not satisfactory. The research reported in this paper departs from this point, and focuses the potential of
ontology-based techniques to enable knowledge use about metadata domains to provide enhanced behaviour
and improved performance.

Recent studies have attempted to unify different learning object definitions (McGreal, 2004). These
studies show that learning-oriented entities in a repository have a high variability on its characterisations. The
non-existence of a common vocabulary, as well as the coexistence of different learning object definitions,
point out the need of flexible repositories that can fit all existent conceptualisations.

Formal ontology as a discipline (Welty and Guarino, 2001) is aimed at studying possibilia, so that it can
be used to compare learning element representations according to the flexibility of their coverage, and term
subsumption properties. In fact, ontological representations can play an important role as support for sound
semantic models that fulfil a number of new requirements related to automation, such as search, retrieval or
composition of new learning materials from others that already exist. The existence of ontology-based
schemas becomes essential when some of the functions are to be delegated to automated or semi-automated
systems, following the semantic web vision (Berners-Lee, Hendler and Lassila, 2001).

This paper shows the progress of a research in course whose main goal is the design of an ontology
schema capable to both bringing more flexibility to the description of the entities stored in SLOR and
allowing automated functions and/or task delegation to agents. The most significant contributions of this work
are the following:
1. Semantic description of standard IEEE LOM1 metadata.
2. Resource description with several ontologies.
3. Design of the SLOR core basic functions that enable a variety of ontological characterisations about
the learning object concept.
4. Upgrade searching and browsing functions.
5. Generalisation of the concept of ‘semantic repository’.

Those contributions are the starting point to others applications that will eventually make use of the
proposed technique. In the rest of this paper, learning object repositories lacks are outlined. Later on, some of
the benefits of the so-called semantic repositories are exposed, as well as some problems found during the
SLOR prototype development. Finally, conclusions and further work are provided.
2. Lacks of current repositories
Currently, learning object repositories, such as MERLOT or CAREO, describe web-oriented learning
resources by storing metadata records linked to them (in a general manner). Thus, they guarantee best search
results on the basis of a fixed structure of the knowledge stored. Nevertheless, better searching capabilities are
not the only advantage provided by these repositories: others, like cooperative review work about learning
objects are definitely remarkable, since the quality of the content can be this way reviewed, analysed and
discussed by several users of the repository.

It is expected that these repositories can play an important role about e-learning in the near future. Both
humans and software agents (such as Learning Management Systems – LMS) will be capable of querying and
searching the information stored in them. However, to execute some reasoning or inference tasks, quality
metadata information are extremely necessary. For the purpose of this work, ‘quality metadata information’ is
defined as the information that has accomplished, at least, minimum requisites of formal description and
where the data provided are based on a pre-established, uniform and preferably universal formal schema.
Current repositories lack a conceptual model establishing what a learning object is and what its metadata
descriptors are, as this is often related to each of the many different conceptualisations available. However,
without a universal agreement about the metadata model to be used and no certainty about the requirements of
an utter formal description, there is an important shortage that complicates the automation of management
tasks in these repositories. Currently, the quality of metadata records relies on the following factors (among
others):
• The learning object creator’s willingness to associate metadata information at the time of adding
their materials to the repository.

1 http://ltsc.ieee.org/wg12/par1484-12-1.html