Who said that? Metadata, trust, and the Semantic Web
Scott Wilson, CETIS staff
May 13, 2002

image:Who said that? Metadata, trust, and the Semantic Web

A new paper from researchers in Stockholm looks to the future of metadata in eLearning, and in particular the next-generation Semantic Web proposed by Tim Berners-Lee and others. But to get there we need to past some of our preconceptions about metadata - like the idea that learning object metadata can ever be objective.

The Knowledge Management Research (KMR) group at the Royal Institute of Technology aren't just talking about the next generation of web technology; they've started to build it. Right now they are working on (amongst other things) Conzilla, for example, a semantic browser that works on a conceptual model of knowledge, and Edutella, a peer-to-peer metadata sharing framework (somewhat akin to Napster).

So when a paper comes out of KMR entitled "Semantic Web Metadata for e-Learning - Some Architectural Guidelines" it's wise to sit up and take notice.

Metadata is a much abused term these days - particularly in e-Learning. Metadata was the first specification to come out of IMS, and one of the most widely adopted. However, KMR are keen to point out some of the key misconceptions about metadata that are, in their view, holding back the Semantic Web vision.

One of the key misconceptions is the notion of the objectivity of metadata. One of the basic models assumed by most e-learning technologies is the idea of a resource with its associated metadata record; for example, an IMS Content Package manifest with an IEEE LOM metadata record present for each object.

This model presumes that the resource in question can be objectively described by that metadata record; yet in our daily lives this idea of there being an objective description of things like films, books and so on does not exist there are only opinions and interpretations. The kind of metadata being presented as "objective" with a learning object may be no more unbiased and reliable than the publishers blurb on the back of a novel or DVD movie.

Looking at it this way, metadata is transformed from a description of the object to an interpretation provided by a particular source: "When metadata descriptions are instead properly annotated with their source, creating metadata is no longer a question of finding the authoritative description of a resource. Multiple, even conflicting descriptions can co-exist".

For the semantic web, this issue is dealt with using the inbuilt meta-metadata support in technologies such as RDF. Using RDF it becomes possible, for example, to search for resource only using trusted sources of description (like most people I usually choose which movies to see based on reviews in magazines whose opinions I trust). When you discard the notion of Absolute Truth in the matter of metadata, trust becomes the key factor.

These issues inform KMR's discussion of metadata architecture - envisioned as an "ecosystem" of ever-evolving interpretations layered upon objects as they are created, used and reused. Instead of metadata only being created and published with an object, metadata is instead constantly added, edited, and appended by diverse sources over time.

There's much more of course - this is an important paper. We have the debate on RDF vs XML for constructing metadata. There is useful discussion on how metadata is used, above and beyond "a digital replacement for library indexing systems". We also have some useful techniques and tools for getting the ecosystem up and running.

Read this paper online at the WWW 2002 site; then try diving into the KMR website. This is new and interesting!