Thursday, 4 November 2004

Learning Object Metadata (The Buntine Oration - Reflection 3)

This is the third post in a series of reflections triggered by Stephen Downes's Buntine Oration. The initiating one reflected on the background of my being here today. The first real reflection was on federated search. The second is on learning object. This post is on metadata.

As I have previously disclosed, I was lucky enough to have exposure to IEEE LOM at its early stages, having spent some time in harmonising the earlier versions with other efforts. However, also at about the same time, I have developed my own "3-type data model" in conceptualising data about data, which is still an on-going project, see here and here.

Two metaphors at the early stages in my exposure to metadata were very compelling:

  • Metadata is like the label on a can. Without the label, we have to open up each can in order to find out what is in the can. So, for some years, I was always carrying two cans when I went to conferences and talked about metadata.

  • The notion of "author", "writer" or "creator" all carry similar meaning. But software will have great difficulty in recognising that these are essential the same term. So I was convinced that a metadata standard was necessary in order to simplify the intelligence behind the software in order to "harmonise" all the efforts around the world at that time.

  • My view gradually changed.
  • First of these was the "3-type data model" I was developing. I realised that "metadata" is not the only way of providing the mechanism for support of discovery of learning resources. In fact, the commercial search engines have demonstrated that an inverted index type of "type 2" data is more efficient and scalable. The use of metadata as part of the resource discovery mechanism became less and less appealing to me.

  • As my job still required me to improve the metadata concept, I also looked at the fundamentals of metadata: the process of involving human to categorize information is full of flaws, inconsistency and laden with multiple hidden agenda. Harvesting was investigated. Collaboration between subject gateways studied. But I became even more convinced that we needed a better mechanism. [I remember reading something about metadata crap lately in the blogosphere. If anyone can point to us where it was posted, it will be highly appreciated.]

  • As part of another project in studying the inter-operability of learning resources from different learning strategies, see this and this, I realised that learning resources used in one learning strategy were hardly usable in another learning strategy without re-work, re-word or changes. That was not the kind of interoperability I was envisaging and was promised. I also realised that the ability to "discover" learning resource was a lesser problem than the ability to "render" resources in meaningful ways under the learning paradigm that a particular academic (or faculty in US terms) is subscribed to. We were addressing a problem from the wrong end!

  • The last straw that broke my belief in metadata came from the concept of "dumb down". At that stage, we realised that it was difficult to have a single set of metadata elements to satisfy all needs. Hence, the qualifiers were introduced. We had qualifiers for the element itself - either extending or narrowing the semantics associated with a particular element. We also have qualifiers for the restricted values assignable to these elements in order to cater for different needs and community background. In order to enable interoperability, the concept of "dumb down" was needed. When a cross-walk was unable to match the exact semantics of the element, we "dumb down" to a broader, agreed element. Good in a sense, but, hey all the hard work that gone into creating the differences are all gone!

  • The metadata effort drew a lot of experiences from people with experience in categorising information - the librarians. However, I believe the effort has put less focus in drawing the experience from the "chalk face". The metadata effort has attracted a lot of interest from technology companies who provided solution of content management and from publishers who wanted to maximize the use of their content. But I still failed to see how metadata can help the chalk face in their daily work. Most academic are expert in their own discipline. They have a very good knowledge of the kind of resources they can and would use. They don't need resource discovery support. They need tools to help them use the resources they already own!

    Just another point before I finish off this reflection. I thought Stephen and I have differences in "federated search". After a few comments exchange, I realise that we are actually have the same point of view. From a technical point of view, I don't see any reason why we should not support "mega-searching", i.e. allow searching across multiple resource repositories (assuming that there is such a need). However, I agree with Stephen completely that encouraging "lock-up" content, either by IP enforcement or depositing into deep repository is doing no good for education.

    No comments: