Regresar

Populating learning object repositories with hidden internal quality information

Abstract:

It is known that current Learning Object Repositories adopt strategies for quality assessment of their resources that rely on the impressions of quality given by the members of the repository community. Although this strategy can be considered effective at some extent, the number of resources inside repositories tends to increase more rapidly than the number of evaluations given by this community, thus leaving several resources of the repository without any quality assessment. The present work describes the results of an experiment for automatically generate quality information about learning resources inside repositories through the use of Artificial Neural Networks models. We were able to generate models for classifying resources between good and not-good with accuracies that vary from 50% to 80% depending on the given subset. The preliminary results found here point out the feasibility of such approach and can be used as a starting point for the pursuit of automatically generation of internal quality information about resources inside repositories.