We'll have to work it out in practice. But yes, in actual formal
modelling so you know what your data means for use, I think it's pretty
clear you need to know entities are being described, not just a random
collection of free form "attributes".
Perhaps a particular film community 'work' corresponds to a particular
general library community 'expression'. And that correspondence can be
expressed, in SKOS or any other schema you want to express it in.
Perhaps the film community even uses a slightly different formal model
than the one used by the general community, with extra attributes, or
attributes on different entities, or even extra entities.
But either way, if I'm interpreting data, it's not enough to know that
_something_ is "black and white", I need to know _what_ is being
described as black and white, and I need the formal model the agent that
generated the data used, so I know what they thought they were
describing. If the film community has a different formal model than the
generalized one, then so be it, but I still need formalized models
(perhaps more than one), and I still need to know what the entities
being described are.
And the only way to figure out how this is going to work in practice is
to start doing it.
Jonathan
Karen Coyle wrote:
> Jonathan Rochkind wrote:
>
>> Yes, I agree. So let's start moving on it. Note well though that
>> "strongly defined data elements" implies you know what _entity_ those
>> data elements apply to, which implies that you've formally defined
>> some entities.
>>
>>
>
> I'm not sure that's the case with WEMI, which is why I have trouble with
> them as entities. Let's take the example of "color" (or "colour" as it
> is called in RDA). You have a resource in which color is important. You
> code it ("b&w", "colored", "red"). For a film, "standard" RDA puts color
> in the E. Film librarians are saying they will put color in the W (so a
> colorized film becomes a different Work). Meanwhile, you have a
> manifestion that comes out with a particular color, and your library
> considers that an attribute of M.
>
> Do you need to define "color of W", "color of E" and "color of M"? If
> you do, then the film archive's record won't match the RDA record. If
> instead you have an element "primary color of visual aspect of resource"
> (or something like that), then they do.
>
> Believe me, this question of how much you need to tie an element to an
> entity is not an easy one to answer. We struggled with this a lot in
> interpreting the RDA elements for the Metadata Registry. I'm fine with
> "Work title" v. "Expression title" -- those make sense to me. But for
> other elements, like color, it doesn't seem to work as well. This is
> where WEMI breaks down for me, because it seems to force elements into
> boxes artifically. Is it because it mixes up the conceptual
> relationships (expression of a work) and physical description? And I'm
> not sure what to do with Subject as an *entity*.
>
> All of this being why I would prefer to have a "Resource" entity that
> has content and carrier, and a lot of relationships like "is expression
> of" and "is translation of" rather that trying to fit things like
> "color" into a single box.
>
> kc
>
>
Received on Wed Oct 21 2009 - 12:45:54 EDT