Current repositories’ shortages

Currently learning object repositories, such as MERLOT or CAREO, describe several web learning resources storing metadata registers linked to learning objects (in a general manner). Thus they give a guarantee for best search structure on the knowledge stored. Nevertheless, the searching is not the only advantage provides by this repositories, since they provide cooperative review work about learning objects, therefore the quality of the content is reviewed, analyzed and discussed by several repository’s users that gain access through internet.

These repositories play a new important role about e-learning future. Both humans and software agents (such LMS) can query and search information on these. It is necessary quality information within metadata records in order to execute some reasoning or inference tasks. Understanding “quality metadata information” as information that it has accomplished minimum requisites of formal description, also the data provided have based on a formal schema pre-establish, uniform and universal (preferably). These repositories lack of a conceptual model that establishes what a learning object is and what metadata descriptors are there associated to each one of different conceptualizations. However without a universal agreement about a metadata model to use and neither certainty about the requirements of an utter formal description, there is a heavy shortage that it complicates the automation of these repositories. Currently the metadata records quality rely on following factors (among others):

- The learning object creator’s willingness to associate metadata information at the time of adding their materials to the repository.
- The editing capabilities or tools provided by the repository for learning object crea-tors to associate metadata information to their materials. These capabilities heavily restrict the kind of information that can be stored and its level of formalization.
- The level of cognition or instruction of learning object creators on existing metada-ta practices and standards.
- The conceptual model of the repository, that is to say, what do the repository creators understand that a learning object is, and what structure the metadata information associated to it should have.

Writing metadata information about the learning objects, in the form of records de-scribing their content, facilitates a number of processes such as storage, search and retrieval from distributed repositories, as well as the composition of new learning materials as an aggregation of others. Accepted metadata specifications and standards make learning objects interoperable and reusable, but a number of shortcomings regarding current learning object metadata still exist:

- Current standards are purposefully descriptive instead of normative: they are intended to give information about the contents or the format of the learning object, but do not generally entail explicit run-time semantics for Learning Management Systems (LMS) that use the learning object (Sánchez-Alonso & Sicilia, 2005). An exception to this is the IMS Simple Sequencing Specification, which allows representing “the intended behavior of an authored learning experience such that any learning technology system can sequence discrete learning activities in a consistent way that includes explicit runtime support” (IMS, 2003).

- The information in a learning object metadata record is not, as currently defined in international standards such as LOM (LTSC, 2002), machine consumption-oriented. In fact, most metadata in current learning object repositories are no more than an overall content identification and description, thus providing limited value from the viewpoint of delegation. In addition, information in metadata records is mostly in the form of unstructured texts written in natural language, such as faculty member’s peer reviews and comments, a kind of information that software agents would find it difficult to process. All this hinders the possibility of programming applications capable of “behaving” according to the information in a learning object metadata record (for example, to process a sequence of actions into a learning management system according to object metadata information)

All shortages exposed deriving to a software agents’ interpretation problem about the existing elements in a repository. For example, the agents’ knowledge that it can be interchange between others, it is formalized in cognitive structures called ontologies, therefore the use of one ontology (that it provides a formal model of conceptualiza-tion) is the only way to process and understand the knowledge holding within meta-data records.

Other remarkable problems are referred to the communication and interoperability. Because of these repositories haven’t got any restrictions and no interchange schema defined, they are isolated in a close system whose only advantage is the Web interface provided. All knowledge interchange functions between different kinds of repositories can not accomplished, since there is not any agreement and recommendation to fit these types of transactions. Furthermore, not any software agent at all can be accomplished an autonomous treatment by itself, as an example choice a learning object (rely on the clients’ preconditions) between others that have the same didactical objectives. Another example about currently restrictions, is the impossibility of tending between different trust levels or append new schemas to the repositories to store metadata records that hold a planning by cost in accordance with the student‘s knowledge levels (Soto & García, 2005).