deborah: the Library of Congress cataloging numbers for children's literature, technology, and library science (Default)
[personal profile] deborah
Just like the other times I've seen these gentlemen speak, I found this panel -- and Carl Lagoze in particular -- to be less provocative than frustrating. I never get the sense that Lagoze is listening to anyone in the room, even though he pays more lip service than any other panelist I've ever heard to saying 'talk to me, tell me what I'm doing wrong. But he takes all questions about his process as attacks and perforce responds angrily and defensively. Lynch, on the other hand, is friendly, but only wants to communicate with people he knows. It's hard for me to get past the aggressive, defensive mode of communication to see potential value in the ORE effort, but I'm trying.

Overview of OAI-ORE (Van de Sompel):

Problem statement: We need to facilitate use and reuse of compound information objects.

  • metadata contains references, but it's not clear references to what:

    • digital object?

    • authenticatation pages?

    • more metadata?

    • datatype chooser among aggregate bitstreams?

  • this works for humans (thoguh from working with patrons, I'd argue that it doesn't) but not for automated information interchange

Goal: solve in a way that works for general web community, not just libraries and scholarly community (eg both arxiv and flickr).

The master resource contains a number of representations (eg. PDF, RTF, etc. deb sez: think FRBR). Often you can't directly accesses these representations.

  • though there might be some metadata to conceptually package the representations (eg a splash page),

  • this isn't true at a machine stage. on the system itself, these are just distinct, separate URIs

    • question: is he assuming there is no METS-style packaging of the aggregate object?

  • so they publish a named graph ("resource map") which gets its own URI to express the relationships among the elements of the aggregate object

    • question: how is this not replicating the work of existing packaging metadata schemas? (eg METS again)

Anyway, this helps reuse by allowing direct reference to

  • resource map

  • direct bitstream

  • direct bitstream as contextualised in the resource map

JISC (Jacobs):

Jacobs gave a rough overview of what JISC is doing.

OAI-ORE in chemistry (Lagoze):

Lagoze argues that chemistry provides a good pilot example for a variety of reasons, including:

  • publishers and professional societies not committed to open access

  • treatment of expression of actual molecular compounds as compound objects

  • real commitment from scientific community to make this work

  • mellon and microsoft will fund the research, they hope

(deb asks: what about user front ends? it's a false statement to believe that the core problem is one that;s easy for humans.)

Lagoze's discussion points

  • is this an important problem? solid use cases?

  • is the commitment to integration with the general web appropriate? (deb sez: why focus on something non scholarly users won't use. cf. dublin core, RDF, semantic web, etc. so there's no point in keeping commitment to general web unless we can see the selling point for the general community.) (deb sez: as with metadata, the human problem is harder than the technological. describing is wicked hard.)

  • are named graphs the right approach?

  • other ideas?

  • how to interoperate/compete with other interoperability efforts? (deb sez: see my METS question above. deb also sez: compete? WTF? We shouldn't be competing kthxbai.)

  • semantic web too complicated?

  • semantic web and named graphs don't work together yet, so...?

  • can there be a general core which can be qualified with community specifc syntax?

  • trust and provenance?

  • preservation?

Lagoze's answers to the key raised questions, including mine:

This is not intended for user generated metadata. It's a way to describe relationships among aggregate objects, and may even use METS or DC to do so. However, this answer was not necessarily shaed by the other speakers at the podium, and it's not clear to me (sez Deb) that he's solving a problem that needs that much solution. Sure, a more formal language to describe aggregate objects would be nice, but as Lagoze himself admits, these aggregate objects already exist, there just isn't a formal language to talk about them in a machine readable form. Is this really the great metadata problem that deserves all the Mellon money?

He would probably irritate me less if I didn't believe that he thinks that throwing technological solutions at them will solve the Great Metadata Problems of Our Time; I don't believe he fundamentally believes that the GMPoOT is the problem of humans decribing pages, but instead he thinks the perfect technology to solve the GMPoOT just hasn't yet been created.

Ultimately I think they're solving a real but tiny problem, and I wish the effort were being spent on solving realler, bigger problems.

Date: 2007-06-22 04:46 pm (UTC)
From: [identity profile]
It's a small problem with potentially hefty implications, is where I stand. Solving it would make a lot of interesting things possible, and would also kick IR-software developers out of their silos, which I am absolutely all for.

I have publicly wondered ( why OAI-ORE never seems to mention METS.

I also share your misgivings about Lagoze et al. These are the same yutzes who admitted last year that they thought NSDL would "just work," and they're getting OAI-ORE into areas where it just doesn't belong (see above link). If somebody can whip them into line, maybe this will work. If not, welcome to the next OpenURL, minus OpenURL's adoption track record.
Page generated Apr. 18th, 2014 03:35 am
Powered by Dreamwidth Studios