If you don’t have the pleasure of being a metadata geek in your day job then, move along folks, nothing to see here.

For the 9 and 3/4 people still reading this post, this report from Jorum is worth a read, though not the magic bullet you’d hoped for from the title. The report mainly looks at Jorum’s own practices in regards to automating metadata collection for learning resources (sensible enough too, mostly all ones we practice in SOL*R) and near the end surveys 5 other systems out there to consider what lessons are to be learned. Another that could have been included here as potentially useful is Yahoo’s Term Extraction Service, but as is the case of the others they look at, it holds no magic solution.

The report ends with a list of recommendations for the Jorum service, all of which seem very sensible as an approach to incremental improvements. I wish I could say more, but I am in such the same boat that I won’t. Suffice to say the ‘Survivors of LOR’ support group is meeting at my house on Wednesday, new members always welcome. And bring beer, it’s that kind of support group. – SWL

Wow, what can one say, these reports from the Cetis Vocabularies Project are nothing if not exhaustive. Pretty well everything one could ever want to know about controlled vocabularies for pedagogical resources are contained in these three reports. And if you needed more evidence of the width of the gap between the ‘big standards‘ approach to learning resources and where the best minds of the loosely coupled elearning blogosphere seem to be going, well look no further.

I am really in no position to cast stones; I’d be lucky if my house were made out of even glass, and I sure ain’t without sin. But the 121 pages that comprise the first two survey reports, the Pedagogical Vocabularies Review and the Vocabulary Management Technologies Review, seem hardly to justify the tepid 7 page ‘Recommendations’ document that follows. Study study study, disseminate, more study, pilot a bit, repeat. Sorry guys, I wish I could be more enthusiastic about this; I want to take succour in the belief we can control the growing chaos, find sense through old patterns and methods, but you know what, I can’t do it anymore, I have seen the light, and this is not it. Whoa, dude, your unconscious is showing…put on a towel or something, at least! – SWL

Rarely do you get a chance to read reflections by someone with as much experience as Stuart L. Weibel, the Senior Research Scientist with OCLC, on 10 years of work around Dublin Core and digital metadata standards. And frank too – he concedes that we’ve all perhaps been too optimistic about the ‘author contributed’ model of metadata submission, and also that metadata in general has to find its place (if it has one at all) in the full-text searchable digitized world of the web. Check out also the great pictures of the bogeys (wheel carriages) that move entire trains between railway lines of different gauges between China and Mongolia. – SWL

Possibly of interest to some of you metadata mavens out there is this wiki where you can find ongoing work of the Dublin Core Education Working Group, including this early draft of an application profile of Dublin Core to apply to educational materials. It’s not exactly ‘knock your socks off’ kind of stuff but then let’s hope metadata never is considered such.

Clearly an improvement over straight ahead Dublin Core in adding things like ‘Education Level,’ but as some folks from the library community recently reminded us, it’s not like anyone’s actually using much more than the equivalent of Dublin Core anyways when they use the LOM. Oh yeah, I keep forgetting, automatic metadata generation/harvesting is going to save our souls and use all the LOM elements to then rapdily (magically, automatically…) re-assemble personalized learning. LOM mani padme LOM. Boy, am I having a rough Saturday! – SWL

Raymond Yee helpfully points out this great article by Dick Bulterman titled “Is it Time for a Metadata Moratorium” from which Raymond extracts this true nugget:

“For nontext data – such as video, images, audio, and so on – direct mining is difficult, but exactly at the point that metadata might be useful, manual creation simply doesn’t get done because creating useful metadata descriptions (the proverbial thousands of words) is not in the critical path of content creation.”

Seems obvious, right? And yet, hyperbolic testimonials to the wonders of flickr and del.icio.us aside, this seems exactly the problem (one of many) that countless repository initiatives seem destined to replicate in positing a repository ‘container’ that isn’t even loosely coupled with either the content creation or content use/re-use environments or workflows. Mea Culpa. Mea Maxima Culpa. – SWL

This article by Wilbert Kraan of CETIS will undoubtably make the rounds in the next few days, but it seemed too signifcant to not post on myself.

Many of us have heard speechs and read articles by one of the IEEE LOM’s creators, Erik Duval, to the effect that metadata needs to become more invisible to end users, in part through automating its creation. Well Erik and the crew at KU Leuven haven’t just talked about it – as Wilbert describes, they have released an open source ‘framework’ for the automated generation of metadata based on the context of its use. Extremely exciting news. – SWL

It’s been such an insane couple of weeks for me with work and travel that the best I can do is simply point to this new D-Lib article by Jean Godby and friends. But if her last piece for Ariadne surveying application profiles of the LOM was any indication, you should go read this now. – SWL

Score one for the librarians!

This article by Carol Jean Godby of the OCLC is an absolute bombshell and a must-read for folks working with learning object metadata standards. She follows up on works by Norm Friesen and Lorna Campbell that survey existing application profiles of the IEEE LOM with a view to answering three main questions:

1. Which elements are most widely adopted?
2. What are the prospects for interoperability given these profiles and the entirely optional nature of any of the elements
3. What can be learnt about the motivation for developing an application profile (a.k.a. why can’t us educational technologist just submit to one standard way of describing things or let the librarians do it)

Somewhat unsurprisingly, like Friesen and Campbell before her, she reports that the most used fields from the LOM can be easily mapped to the existing Dublin Core fields, and that we’re pretty much all over the map when it comes to all of the special ‘pedagogical’ type fields that were supposedly the motivation for this whole exercise in the first place. (more…)

When it comes to the potential for interoperability delivered by these various application profiles, Godby’s analysis seems to say it’s murky at best. At one point, she sums up “the best prospects for interoperability are local, and they degrade as institutional, linguistic, and cultural boundaries are crossed.” Additionally alarming (though painfully accurate) is her statement that

“since subject data is sparsely represented, and subject classification schemes for learning objects are still under active development, discovery strategies for LOM records will probably be restricted to known-item searching.”

Finally, she lands a real body blow with a seemingly offhand comment:

“Nevertheless, this evidence constitutes only a precondition for interoperability. A study of application profiles must make the simplifying assumption that profiles are interoperable if they recommend the same elements. But … two application profiles might use LOM.Classification.Purpose and still fail to interoperate because this element could be used to annotate different facets of the resource, such as pedagogical intent and position within a knowledge hierarchy.”

Godby’s not out for blood, though, and makes some very reasonable recommendations at the end of the paper, acknowledging that the problem area these standards are supposed to address is itself still very much in flux. The notion of two layers of metadata, one core to maximize interoperability and harvesting, the other with more local data, seems on the surface worthy of further discussion. In any case, this paper and its findings deserve a lengthy response from many in the elearning standards field who have promoted the LOM (and its application profiles) as the way forward for discovering reusable learning content. – SWL

As mentioned by Stephen, this is the second year Norm Friesen has produced this important report. The survey focuses on two questions: 1) “Which elements were selected for use or population?”; and 2) “How were these elements used, or what where the types of values assigned to them?” Maybe this makes sense as a starting point, but it misses out an even more important question for the builders of repository systems and other consumers of metadata – how are the elements that are selected for use actually employed to deliver functionality to users (or other systems). Maybe that is a premature question, but the current situation resembles a comment a fellow from Oxford made in the Learning Design working group at the Alt-I-Lab sessions a few weeks back – “It’s like you’ve spent all this time working on the data model before you had an idea what the application it was supposed to support was going to be.” Don’t laugh, that’s exactly what this feels like at times.

Which is why the last paragraphs of the paper are so important, the ones in which Norm engages the position laid out by Erik Duval and Wayne Hodgins in their paper titled ‘Metadata Matters.‘ I am sympathetic with Norm’s concern that

“speculations on future developments in technologies and the social practices that develop with them are notoriously prone to error, and especially subject to ideological and other distortions. While the possibilities of such developments and solutions should certainly not be ignored in standards development, they do not provide a solid foundation for standardization work.”

But maybe what this means is that metadata standards need to be driven from two directions that are in tension; one from the direction of studies like Norm’s that look at what fields are currently being populated, but then also from the perspective of current and future usages of the metadata fields so that the limitations of manual metadata collection practices (one of the points I think Hodgins and Duval are trying to make) don’t end up dictating a standard that ultimately doesn’t enable any new functionality (which I thought was supposed to be the reason for all of this to begin with.) – SWL