Quoting Paul Houle <[log in to unmask]>:
> I got into the semantic web because of my interest in Dbpedia,
> Freebase and other things that I call "generic databases". The most
> prevalent topics found in generic databases are
>
> * People
> * Places, and
> * Creative Works
>
> which form a kind of "holy trinity".
I like that image :-).
It seems to me that we need to treat metadata the way we do programs
and _modularize_. One problem that I run into, and that Renato
mentions, is that there are concepts embedded in metadata sets that
should exist at a higher level or that should be defined outside that
particular metadata context. For example, the library data model,
FRBR,[1] has defined "event" "place" and "object" as entities that can
be subjects of a bibliographic item. Unfortunately, they are defined
solely as subjects, and, even if they weren't they don't link to
anyone else's use of "place." Yes, as you say, "place" is a common
metadata property.
It feels like we should have more metadata properties that, like DC
terms, are defined at the most general level (e.g. as subproperties of
owl:Thing, or subclasses of ... I guess owl:Class or rdfs:Class -- I'm
not clear on the difference). Then when one is developing metadata
there would be properties whose definition isn't bound to a single
metadata element set or isn't made confusing by that context. The new
metadata set could use these properties or could define
sub-properties/Classes if necessary. There have been some starts at
this, mainly on vocab.org. Perhaps part of the problem is that we need
to ensure *discovery* of those generalized metadata definitions so
that they can be re-used. There needs to be a place to go (see, place
again) to find re-usable metadata properties. I don't find web
searching to give me good results when I'm looking for re-usable
metadata terms.
kc
[1]
http://www.ifla.org/en/publications/functional-requirements-for-bibliographic-records
> In the end, since people have been bulk loading Freebase with
> information about books, TV episodes, music tracks, and such,
> Freebase looks more like a bibliographic database than anything else.
>
> FOAF was designed by academics (just around the time modern
> 'social media' was being born) and probably the most important thing
> you can say about an academic is what they've published. So it's
> pretty natural that a vocabulary for talking about academics is
> going to place a special role on Documents. If we go forward a few
> years to SIOC, the idea of a Document extends more to blog
> comments, tweets, photos on Flickr, etc..
>
> I think it's natural that people are going create vocabularies
> by selecting terms from existing vocabularies. BIBO's a good
> example of this. It picks from dcterms:, foaf: and a few other
> namespaces where they find something appropriate. On the other
> hand, if you think other people's vocabulary isn't quite right,
> you're also free to make your own predicates.
>
> Something that's bugged me, for instance, is that a big part
> of SIOC is about structural metadata. It's very good for going to
> some site like say,
>
> http://jalopnik.com/
>
> and explaining the structure of it. Yet, the type system in
> SIOC is chauvinistic towards the more frivolous dimension of what
> people are doing in social media today... It doesn't have types for
> whatever the 'next big thing' will be two years from now, and it
> doesn't have the right kind of types to describe, say, a PhD
> thesis published online or a site like
>
> http://shakespeare.mit.edu/
>
> Whenever we create vocabularies that are intended to be shared,
> we're always tempted to "close the world" to make the problem
> manageable, but we're always confounded by an "open world" that
> won't fit in any specific box.
>
--
Karen Coyle
[log in to unmask] http://kcoyle.net
ph: 1-510-540-7596
m: 1-510-435-8234
skype: kcoylenet
|