Learning Object Metadata use survey: sticking the short and wide in the long and thin
The recently revamped CanCore initiative, with assistance from the Finnish delegation to ISO SC36, completed a survey of the widely used IEEE standard for Learning Object Metadata (LOM). The findings paint a picture of communities using relatively small parts of the long list of elements that the LOM provides, but do use their own vocabularies for the elements that are used.
The survey was done for the [deep breath] working group 4 (WG4) "Management and Delivery", of the International Standards Organisation (ISO) Joint Technical Committee 1 (JTC1) Sub Committee 36 (SC36), "IT for Learning, Education and Training" standards body. The august ISO is legally the ultimate authority where nation states agree on standards in a painstaking process. WG4 of the SC36 part of ISO is working on a metadata standard of its own, and it was felt that a thorough overview of how the existing standard in the field, the IEEE's LOM, was used in practice.
The survey team had a look at five sets of LOM records from the UK, China, Canada, France and the European Ariadne consortium. The size of the sets varied from 75 to 3000, and 75 records were sampled from each set, and manually examined for the survey. On a global scale, there really is plenty more where that came from, but CanCore's Norm Friessen and Turku University's Lassi Nirhamo have other things to do with the rest of their lives, so this is where the survey stops for now.
Most people who work with the LOM will probably be able to confirm that the results look representative. First, the most frequently entered information in LOM records were about an object's identifier, it's title, description, authors, technical and educational format or type, and (discipline) classification. The least frequently entered information was about duration of the learning activity, it's difficulty, structure, and granularity.
What's notable here is that the popular information maps almost straight across to the simpler, and very widely used Dublin Core metadata element set. That means that those elements which are most the most educational about the LOM, are also amongst the least frequently used. The good thing about this state of affairs is that it is fairly easy to get interoperation from e-learning to the rest of the world.
Other findings include the fact that nearly all sets had custom vocabularies in their profiles of the LOM to express characteristics that are mostly relevant to the community from which they came. This is to be expected for a metadata standard, but there were some difficulties in how these vocabularies fitted into the LOM, and how other people can find out what these vocabulary items mean.
On the technical side, there were some clear difficulties with the peculiar structure of the LOM: it's neither a straightforward list, nor is it a typical 'tree' of elements. The result is that it isn't always straightforward to deal with LOM metadata record structures in simple tools.
Also problematic was the vCard format that is used within the LOM to represent things like authors. The good thing about vCard is that it is near ubiquitous in implementation, and it does a reasonably good job of describing people. Trouble is, not all authors or other contributors to a learning object are people; some are organisations. Also, both the structure of vCard, and the way it is represented sit awkwardly with the XML format in which the typical LOM record is exchanged. The result is that these parts frequently get mangled.
On the more positive side, the rich constructs that the LOM allows is frequently used effectively, and, above all, the diversity of the survey's records and the places they come from shows that the LOM really is a very widely used common structure to describe learning objects.
The survey can be downloaded from the ISO JTC1 SC36 website (704 Kb, PDF).
The CanCore initiative's website has much more information about the LOM generally, include the comprehensive version 2 of the CanCore guidelines.