Author:

Keywords:

learning object metadata

Abstract:

Relying fully on manual effort in generating metadata is not the way to go: we need automation of this process, as much as possible, without losing the “quality” of the metadata. In this paper we report on our experiences with automatic metadata generation. We briefly outline the first version of our framework for automatic metadata generation and then report on some lessons we learned. These lessons resulted in a redesign of the framework. An important aspect in this redesign is the bottom-up strategy, instead of the top- down design we used before. This means that we start with an abstract specification that is then made concrete in implementations. This should allow for interoperability between installations that do (part of) the metadata generation. The redesigned framework is used in a case study that also extends the metadata generation part by allowing for search and retrieval of the generated metadata. We also report on an evaluation experiment that we set up for this case study, which points out that the automatically generated metadata has a similar level of quality than manually generated records present in the ARIADNE repository. Finally we conclude with some future work.