[foaf-dev] Proposal: deprecate pastProject and currentProject

Dan Brickley danbri at danbri.org
Tue Dec 22 10:28:35 CET 2009


Hi Graham,

On Mon, Dec 14, 2009 at 9:51 AM, Graham Klyne <GK-lists at ninebynine.org> wrote:
> Dan Brickley wrote:
>>
>> There's a risk with this stuff of building parallel sets of
>> properties; those that link to pages and those that link to the
>> primary topics of those pages. RDFa biases us towards the former,
>> modelling clarity towards the latters. Reminds me of the XFN mapping
>> discussion too.
>>
>> workplaceHomepage, schoolHomepage are in former category; pastProject,
>> currentProject too on my preferred reading; but I don't have a strong
>> view.
>
> FWIW, I've been working lately with CIDOC CRM [1] [2].
>
> The issue of present, past and distant past relations are captured quite
> neatly byb an event-mediated structure - things (people, places, objects,
> times) are related to things by events.  These events provide a place to
> hang additional information so that properties like pastXXX and currentXXX
> aren't needed.

Yes, I'm sure we've talked about event models before; ABC etc.

( see also http://lists.foaf-project.org/pipermail/foaf-dev/2008-July/009276.html
)

Once you get into describing the world in terms of events, it can seem
like the answer to everything. Since events themselves don't have
changing characteristics. However, reasoning about the state of the
world, given a (possibly incomplete) pile of event descriptions does,
I fear, roughly correspond to the problem of simulating general
purpose 'commonsense' intelligence. If my car starts out blue, is
painted black, passes through a chain of 3 other paintjob-happy
owners, and then rusts terribly, and then explodes in an accident, ...
can we figure out its colour at various points? Probably not. We might
make informed guesses, based on our knowledge of physical reality,
society and the everyday use of the word 'colour' with respect to
cars.

So in the general case I'm not an optimist about mechanistically
derriving descriptions of the current state of a world from
descriptions of events. That said, there *is* a lot of value in
describing events and activities for their own sake. Particularly in
RDF when binary relations are too restrictive to capture everything
you want to say about an object, so you need a 'reified' intermediate
object anyway, events are quite a natural fit.

> So we have clean and flexible modelling;  the downside is a tendency to
> generate very "deep" RDF structures.  AFAICT, it works as well for
> scientific as well as for its roots in cultural heritage

I often encounter CIDOC in research projects, and there were even
meetings years ago even on 'reconciling' the ABC model (
http://metadata.net/harmony/3rdDelosHarmony_Minutes.doc ) with CIDOC
(I'd dropped out of the ABC thing by then). I don't know of so many
real life apps built yet with CIDOC, and I'm not sure if that's a
problem, of if it serves more as a background conceptual model that
can be used to inspire and document working schemas.

> Does any of this have relevance for FOAF?  I don't know, but there have been
> apparently successful efforts to configure other DC-style "flat"
> vocabularies around CRM (e.g. CDWA-Lite/MuseumDat).

Do you have a reference to that?

Re FOAF and events/activities, yes and no. I don't think we'll use
CIDOC directly, but I would like to have clearer annotations on
properties to say when we know things about the potential for their
values to change. I suspect most practical work will be driven by the
need to merge reconcile different RDF graphs loaded into a common
SPARQL store. As RDF stores grow cluttered with different
semi-compatible 'named graphs' I think we'll have more and more need
for table-of-content metadata describing the source, provenance and
associated dates of each set of triples. So I expect to see
event-oriented stuff plugged in there. Also perhaps
http://xmlns.com/wot/0.1/ could do with an update, and allignment with
the foaf+ssl efforts underway, so some scope for talk of signing
events etc too.

There is certainly a lot of interest around describing events and
activities, for example  my colleagues at VU.nl have
http://semanticweb.cs.vu.nl/2009/04/event/ and there is the Atom/JSON
work at http://activitystrea.ms/ which comes more from a social
network data stream perspective. And of course
http://motools.sourceforge.net/event/event.html ...

In the NoTube project we're looking at aggregating event descriptions
to derrive (with or without user intervention) "FOAF++" user profiles.
This goes a bit beyond the topic of the current thread but is
interesting in its own right...

cheers,

Dan

> #g
> --
>
> [1] http://cidoc.ics.forth.gr/
>
> [2] http://erlangen-crm.org/docs/091125/index.html
>
>
>
>


More information about the foaf-dev mailing list