[rdfweb-dev] Relationship Schema Updated

Michael Bauser michael at bauser.com
Fri Mar 19 04:30:04 UTC 2004

Hash: SHA1

Morten Frederiksen wrote:
> On Sunday 14 March 2004 11:12, Michael Bauser wrote:
>>The LITTLE problem with the "label it with my language or theirs"
>>approach is that FOAF doesn't exactly have a way to label "what language
>> and culture I define my networks with". FOAF doesn't even have
>>properties for nationality yet.
> True, but nationality is becoming less important these days.

This news would probably come as a shock to the State Department. They
have a lot of paperwork involved in the concept of nationality.

But, seriously, I wasn't actually trying to use nationality as part of a
kinship system -- I was trying to point out that using FOAF to describe
complex "background stuff" like "the structure of the foaf:Person's
native society's kinship system" may be an unreasonable expectation,
given that FOAF doesn't even deal with fairly basic "background stuff"
like foaf:Person's nationality. (By "basic", I mean only that the
question "What nationality is this person?" *usually* has a one-word
answer, and it gets written down by the government a lot.)

In fact, it seems to me that FOAF (and its related vocabularies) spend a
lot of time dealing with the *locational* triangulation (who is working
with whom, who is living with whom, etc.), but no time dealing with
*ideational* triangulation. (How the creator of a FOAF file defines
"working", for example, is a significant issue -- does rel:worksWith
require that the related people work for money?) Until and if the RDF
vocabularies used to describe people deal with some of those cultural
assumptions (using RDF logic where they can), *all* such vocabularies
have hidden cultural biases that limit their utility in multicultural

My biggest concern with how the relationship vocabulary handles kinship
is that it builds in two big "21st-Century American" biases (one about
the structural equivilences of certain social relationships, and another
about the relative importance of linear and collateral kinship) that
would force people using it to "throw away" a lot of interesting (even
to themselves) data *and* that the vocabulary doesn't seem to realize
that its culturally biased. (If somebody needed to create a vocabulary
for "American families" with all the biases *explicit*, I'd say "go for
it" -- such vocabularies have their uses, too.)

This is especially problematic (to me) at this relatively early stage in
FOAF's (and even RDF's) development, because a culturally-biased
vocabulary sets a bad example for vocabularies that follow. The
"nightmare scenario" is one where there are a dozens (even hundreds or
thousands) of very narrowly-bounded vocabularies trying to describe the
same phenomena. (Try to picture this: You're defining the kinship
vocabulary for whatever society you live in, and you have to figure out
the owl:equivilentProperty relationships for a hundred other systems'
versions of "cousin".) That's going to make life hell for the software
guys when they try to deal with kinship in any meaningful manner. Making
life hell for programmers is a good way to *not* get a standard implemented.

The system I'm arguing for is an attempt to separate the "storage
vocabulary" from the "presentation vocabulary". Use the anthropological
kinship-mappings to *store* data (which, from some users' perspectives,
may involve storing a little "extra" data), and the culturally-bounded
vocabularies can be used to translate the data into an appropriate
cultural viewpoint *when needed*. A good storage vocabulary could
greatly simplify life for programmers (and vocabuary creators) by
changing the problem from "How does vocabulary A relate to vocabularies
B, C, D, E, F, and G?" to "How doe the storage vocabulary relate to this
presentation vocabulary?". The anthropological vocabulary provides a
good translation key between different cultural vocabularies.

> However, we do have inkel's speak/reads/writes schema [1], and
combined with
> language specific rdfs:label's, it should go a long way. As I
understand it,
> the "correct" labels are more a function of language (and thus
culture) than
> nationality.

I'm already aware of that one, and given nothing else in the data, it
would a *reasonable* inference to say "this person's primary language is
Chinese, so he probably uses Chinese terminology to describe his kinship
network". But it's an imperfect inference, in that two societies can use
the same language while assigned kinship terms differently, *and*
terminologies can evolve through time. Language is definitely *part* of
the equation for determining the native view of a kinship system, but a
really good equation would need more data.

I am, above all, willing to accept that there are differing degrees of
"dependibility" when making inferences from RDF data. Knowing someone's
native language makes inferences about their kinship system more
dependable than *not* having any information, but having a vocabulary
that explicitly described native systems would be better. Right now,
though, I want to concentrate on the "storage vocabulary" part of the
equation, because using RDF to describe systems sounds like a too big a
project for this stage in RDF's history.

Version: GnuPG v1.2.1 (MingW32)


More information about the foaf-dev mailing list