W3C home > Mailing lists > Public > semantic-web@w3.org > June 2020

Re: defining the semantics of lists

From: Thomas Passin <tpassin@tompassin.net>
Date: Thu, 4 Jun 2020 08:41:27 -0400
To: hans.teijgeler@quicknet.nl, semantic-web@w3.org
Cc: 'thomas lörtsch' <tl@rat.io>, 'Jiří Procházka' <ojirio@gmail.com>, 'Hugh Glaser' <hugh@glasers.org>, 'Peter Patel-Schneider' <pfpschneider@gmail.com>
Message-ID: <c5079342-80c2-dc11-f6b8-8c36ef20e346@tompassin.net>
On 6/4/2020 3:23 AM, hans.teijgeler@quicknet.nl wrote:
> We need n-ary relations as described in 2006 by Natasha Noy and Alan Rector in https://www.w3.org/TR/swbp-n-aryRelations/
> Perhaps they see that now as a youthful indiscretion, but the standard we work on (ISO 15926-7) is based on that document.
> 
> These n-ary relations need to be classes and for each relation involved we need to be able to define the role of the object.
> All I have ever heard that such is impossible in OWL, never seen evidence and it renders OWL as less relevant. But we use RDF anyway.

Topic Maps actually do model relationships this way: a relation (called 
an "association" in Topic Maps) relates actors playing roles in the 
association.  Each object of the association has - or rather *is* - a 
role, and that role has a type (or class, if you prefer) and an actor.

This works well, and is probably just what you want.  But when you work 
with them, it's a little annoying to have to go through the machinery of 
finding roles by their type, and then the actor attached to that.

Any topic map can be transformed to an RDF graph, just like any SQL 
database can be converted to an RDF graph.  But the complexity will 
still be there - it is inherent in the facts being modeled.  And if you 
work around the complexity for convenience, you still have all those 
details to handle, except that now they will be implicit instead of 
being explicit.

In the cases you are talking about, it seems to me that you could model 
the (n-ary) relations in the Topic Map way by having the relationship 
objects be the roles rather than individuals, and the roles would have 
(that is, be related to) both a class and an individual instance of the 
class.  This should be OWL-friendly, I would think.
Received on Thursday, 4 June 2020 12:41:49 UTC

This archive was generated by hypermail 2.4.0 : Thursday, 24 March 2022 20:42:10 UTC