[rdfweb-dev] Another relationships proposal

David Menendez zednenem at psualum.com
Sat Mar 20 00:38:49 UTC 2004


Based on what I've written below, I think it would be useful to do a
requirements/use cases phase before developing the schema further. What
information are we trying to capture? How do we envision it being used?
Any vocabulary is going to be too complex for some purposes and/or not
complex enough for others, but I think we can hit an 80/20 point
somewhere.

Ian Davis writes:

> 
> On Friday, March 19, 2004, 8:58:57 PM, David Menendez wrote:
> > True, but the vocabulary is merely asserting the existence of a
> > relationship (and again, { A ancestor B } means "B is an ancestor
> > of A according to the rules A uses"). I think that even the most
> > patrilineal society would accept that people *have* maternal
> > grandmothers, even if they aren't important in terms of inheritance.
> 
> > That is to say, if I say { A ancestor B }, I'm claiming that there
> > is *some* line of descent from B to A.

> I don't think I expressed my point very clearly. What I meant to say
> was that I though the parent/ancestor subclass/superclass relationship
> was not universally shared. See my follow up message where I move
> ancestor to be a peer class of parent.

I think you mean "property" here, not "class".

I don't think we should get too hung up on the actual words being used
to name these RDF terms. As I was using them, "ancestor" and
"descendent" refer to lineal relatives (that is, related by a direct
line of descent). This is orthogonal to whether a society is patrilineal
or matrilineal or something else.

That being said, I guess you could argue that an adoptive parent is not
an ancestor. (Do adopted children consider their adoptive parents'
parents grandparents? I think (without evidence) that enough of them do
to justify the use of "ancestor" here.)

> I think there is another problem with your proposal (and I think this
> is part of Michael's argument). You define mother to be "whatever
> mother means in the culture of the subject". In some cultures the
> sisters of the biological mother are also called mothers. Under your
> proposal (and the current relationship vocab) there is no way to
> distinguish the two because don't know the culture of the subject.

Again, we shouldn't conflate the natural-language terms people use to
refer to their relatives with the RDF terms they might use to describe
their relationship.

In the case where a person addresses his or her biological mother and
her sisters as "Mother", is it really the case that no distinction is
being made? I imagine the women are not interchangeable, since the
child's father is presumably involved with only one of them.

There is a need for some ambiguity in the meaning of "mother", because
it should be able to cover adoption, foster parenting, same-sex couples,
and as-yet-unforseen developments, but the sense of "female social
parent" is pretty straightforward.

I'd like to hear Michael's input on this point, since it was his
comments that lead to my suggestion in the first place.

> Where it causes problems is if the enquirer expects the results to
> conform to their personal cultural rules.

It would cause problems for the enquirer, certainly. :)

That being said, no controlled vocabulary can be all things to all
people.

> This suggests another way to model this:
<snip>
> Here the relationships are not to the subject _:ian but to the family
> unit. This is a strawman easily demolished but it might shed light on
> other models.

I'm not sure that gains us anything in expressive power, but I could be
convinced otherwise.

(thinking aloud...)

I guess this graph:

  [ a Family
  ; father A
  ; son B
  ; daughter C
  ].

implies that B and C are siblings, whereas this does not:

  [ a Family
  ; father A
  ; son B
  ].
  
  [ a Family
  ; father A
  ; daughter C
  ].

(It also doesn't deny that they are siblings, but that's another
matter.)
-- 
David Menendez <zednenem at psualum.com> <http://www.eyrie.org/~zednenem/>



More information about the foaf-dev mailing list