November 30, 2006
Hodgins ponders the future of metadata, while revising the past and present
Apparently, the rise of RSS, blogging and social software is actually the rest of the world finally realising how valuable and successful LOM and LO's have been. At least that's one of the messages I'm getting from Wayne Hodgins' post on the Future of Metadata.
Is it just me that sees pretty much the whole enterprise of LOs and LOM as solving entirely the wrong problems in all the wrong ways, and thank heavens the web had an alternative developing ready to get us out of the mess we'd made for ourselves? Even David Wiley has pretty much abandoned LOs in favour of concentrating on the key principle of openness.
There is some good stuff in here on future work, but to be honest it would be better off ditching most of the legacy rather than trying to convince us of what a wonderful success its been, and all this new stuff is logical progression. After all, anyone talking about "secondary metadata" or "usage metadata" a few years ago was being scoffed at by the LOMerati. What we needed was more, even higher quality, expert-produced metadata, whatever the cost!
So a bit of revisionism mixed in with an overly optimistic view of what we achieved in terms of value realized from the LO/LOM approach, but some nice ideas. For example, while I'm a fan of Pandora and the Music Genome Project (its a fun distraction), the idea of concentrating on musical forms and ignoring things like image and timing is to fundamentally misunderstand pop music, and its results are probably not far off random serendipity. So I'm not sure analyzing the "eLearning genome" would necessarily generate the intended outcomes.
So, what is the future of metadata? Well, yes it should include "attention metadata" (or secondary metadata as we've been calling it for ages), probably some automatic generation (which we've been doing for ages - no-one generates those RSS timestamps by hand, you know), making things like LOM invisible to users (and probably vanish from the back-end too if it continues to do nothing of practical benefit).
However, the 'elephant' of metadata - full text indexing - is noticeably absent from this future. And yet this is the great success story of resource discovery on the web! Metadata is better used for browse lists rather than query, which is hardly the personalization target aimed for here.
(On personalization, I had an interesting discussion with someone recently who is seriously investigating how the porn industry successfully manages this via a process of iterative profile building via click tracking with very limited vocabularies.)
I also don't buy competency metadata yet. As we've discovered in TenCompetence and XCRI this is an absolute quagmire, and we can achieve more of practical benefit with reasonable effort by using latent semantic analysis and regular text labels than using ontologies and schemas to try to pin down human capability in all its dimensions. Hodgins' idea of a "Skill Object" simply reiterates the fundamental flaws of the monolithic LOM "describe everything in one model" approach on an even bigger scale.
Despite all my ranting above (which only shows I'm sad enough to actually care about metadata!), this is a post, and a set of slides, that is well worth reading.