Learning Object Metadata and Automatic Processes: Issues and Perspectives

Generation of learning object metadata • Understanding the issues related with the generation of learning object metadata. • Identifying the opportunities and drawbacks of using automatic techniques for generating learning object metadata. Validation of learning object metadata • Understanding the validation of learning object metadata. • Identifying the opportunities and drawbacks of automatic techniques for validating learning object metadata. Learning object retrieval • Understanding the retrieval of learning objects using learning object metadata. • Identifying the opportunities and drawbacks of automatic query generation for enabling semantically rich retrieval of learning objects. Use of learning object metadata • Visualizing the impact of automatic processes on the practical use of learning object metadata. Executive Summary Reuse of learning material has recently become a leitmotiv for research on computer-aided education. The most obvious motivation is the economic interest of reusing learning material instead of repeatedly authoring it. Other motivations can be found in the pedagogical area since learner-centric teaching theories invite instructors to use a wide variety of didactic material. Since sharing and retrieving learning material is a basic requirement to ease learning material reuse, it is not surprising to see the research community specially focusing on these topics. Learning material retrieval should not only stand on common document characteristics-like DublinCore Metadata Intitiative (DCMI, 2005)-in order to be pedagogically relevant, but also on specific educational data. The Learning Object Metadata (LOM) standard includes such data. Consequently, Learning Object Repositories typically use this metadata for storage and retrieval of learning objects. However, creating a LOM document means to instantiate the almost 60 metadata attributes of the IEEE LTSC LOM specification (LOM, 2005). Such a fastidious task is not compatible with making learning material sharing a customary activity for regular teachers. Therefore, several researchers seriously focus on the metadata generation issue (Downes, 2004; Duval et al., 2004; Simon et al., 2004). The main difficulty for generating LOM documents stands in the educational part of the metadata. Educational information is generally implicit in the learning objects. Therefore, existing metadata extraction methods based on content analysis cannot totally serve LOM generation. The same situation occurs with learning material retrieval. Retrieval effectiveness depends on LOM-based query precision. Consequently, generating effective queries could rapidly be as complex as generating LOM documents. In such a context, the future of learning object repositories will definitely depend on the ability of current systems to facilitate the generation of metadata as well as …