Re: What if two resources are the same subject?

On Tue, 13 Apr 2004, Jan Algermissen wrote:

>Jon Hanna wrote:
>>
>> Quoting Jan Algermissen <jalgermissen@topicmapping.com>:
>>
>> > My first question is what happens when RDF statements from both camps are
>> > combined
>> > and people start noticing that there are two resources being the same thing
>> > (Canada).
>>
>> Well there is only one resource, but multiple URIs,
>
>Ah, that I did not know. So resources may 'span across authorities', yes?
>A single resource can live on different hosts controlled by different
>authorities, yes?

Hmm. Some resources are actually abstract concepts, for which different
authorities provide an identifier. "timbl", "the director", "Sir Tim",
"client number E7t98756" might all be references to one person, used by
different autorities. But perhaps none of them own the actual person under
discussion, and he might not live in the hosts as such.

In other cases, because RDF uses URIs as identifiers, you might mean the
thing that is dereferenced with a particular URI. The notion of "sameness" is
in some senses outside the Semantic Web specifications, which merely define
useful assumptions or axioms that should hold if you declare two resources as
the same thing.

>> Of course sometimes there can be subtle differences (e.g. "Ireland" the island
>> vs. "Ireland" the country) and sometimes those differences can be contentious
>> (e.g. Ireland again).
>
>Ok, agreed (I meant the exact same subject 'Canada').

Presumably that is the territory of Canada as defined in the current
Constitution, and not just the collected citizens, excluding resident aliens,
or some other idea of what Canada means... This is the contention Jon was
alluding to. It is clearer in the case of Ireland, because many people think
to ask if you mean teh island, or the republic of Ireland that doesn't
include the English counties of northern Ireland. But sometimes when talking
about Canada one includes the little bits of territory there that belong to
other countries, and other times one doesn't. For most purposes it doesn't
matter, but being clearer about meaning (where "this is a generic and
ill-defined idea of Canada" is being clear about the limitations of precision
and therefore about the meaning) is valuable in reducing later confusion.

>> You do need to "know" that they two URIs identify the same resource. Sometimes
>> this will be published somewhere (ideally either or both URIs when dereferenced
>> would reveal this fact). It might not be that easy to determine though, it
>> might even be secret (Superman owl:sameIndividualAs Clark Kent, but he doesn't
>> want anyone to know :)
>
>Yes. My concern only was if such knowledge could be accumulated inside the SW
>to increase its overall knowledge.

So I hope the answer is clearly "yes it can, but how carefully it is done
may affect the long-term usefulness of each individual inclusion". We can see
the problems when we find inconsistencie.

The Semantic Web itself is now capable of identifying lots of types of
inconsistency, but can't often tell us how to deal with them yet - that's the
trust management bit that people are working on in various places.

Cheers

Chaals

Received on Tuesday, 13 April 2004 14:29:13 UTC