W3C home > Mailing lists > Public > public-swbp-wg@w3.org > July 2004

[ALL] review of n-ary relations

From: Jeremy Carroll <jjc@hplb.hpl.hp.com>
Date: Tue, 06 Jul 2004 11:08:04 +0100
Message-ID: <40EA7A04.3070008@hplb.hpl.hp.com>
To: SWBPD <public-swbp-wg@w3.org>


Summary: ready to publish, a few comments ...

While I have some comments, none are I believe showstoppers, although I 
think I should draw three comments to the WG attention in case others 
would like them addressed before first publication.

These are:
[A] use of skolemization instead of blank nodes
[B] use of amazon.com instead of books.example.com
[C] references not in W3C house style


I have a number of other comments below, which can be treated as 
editorially.

Expanding on these three points.

[A] use of skolemization instead of blank nodes
===============================================

In all the examples the instances of the relationships are given URIs 
such as (in N3)

:Diagnosis_Relation_1


This has two problems:
1) Introducing a URI node for a concept which is really only an 
existential introduces false negatives in entailment.

e.g.

:Christine      a       :Person ;      :has_diagnosis 
:Diagnosis_Relation_1 .

:Diagnosis_Relation_1      a       :Diagnosis_Relation ; 
:diagnosis_probability :HIGH;      :diagnosis_value :Breast_Cancer .

neither entails nor is entailed by

:Christine      a       :Person ;      :has_diagnosis 
:Diagnosis_Relation_2 .

:Diagnosis_Relation_2      a       :Diagnosis_Relation ; 
:diagnosis_probability :HIGH;      :diagnosis_value :Breast_Cancer .

Despite the fact that the two have identical intention.

Replacing the skolemized name ":Diagnosis_Relation_1" with a blank node 
"_:Diagnosis_Relation_1" resolves this.



2) it is in the same namespace as the ontology, see editorial notes below.


[B] use of amazon.com instead of books.example.com
===================================================

Personally I prefer avoiding any reference and any URL to something 
which is not an example.{com,org,net,edu} URL see RFC 2606, 
http://www.faqs.org/rfcs/rfc2606.html, except where necessary (e.g. 
w3.org URLs).

The text refers to amazon.com whereas say books.example.com would be 
just as intelligible and not be open to misinterpretation as endorsing 
one company over another.

(However, I do not believe this is W3C policy, unless Ralph corrects me)

Further examples of URIs I don't like are in the example files where 
URIs starting http://protege.stanford.edu/ could be seen as endorsing 
protege rather than other tools. I think a w3.org URI would be more 
appropriate, in WG space, for these WG examples.


[C] references not in W3C house style
=====================================

The references section at the end of the doc is very minimalistic. In my 
view this is acceptable in a first WD, but at some point it will need to 
be converted to the W3C house style (see any Rec for an example, best to 
go for a recent Rec)





Further comments
================
[[ *NONE* of my comments need to be addressed for first publish WD. 
However I would like the TF and editors to consider these comments in 
good time ]]


I found it slightly disappointing that two out of the three examples 
were medical. I think it would be easy to replace example 2 by an 
example from a different domain than example 1 and example 3.

RDF reification: I think the sentence referring the interested reader to 
  the discussion of reification in RDF could be deleted. However, others 
in the WG may disagree, and if the editors were to follow my preference, 
it would be worth drawing WGs attention to this change.

The examples are in the turtle subset of N3 (I think) however this 
should be made explicit, with a reference to N3.

The links to the "RDF/XML" code should say
"RDF/XML" and not "RDF/XML abbrev"
(I perhaps flatter myself that the reason you have written "RDF/XML 
abbrev" is following the name of the RDF/XML writer that I wrote for 
Jena. This name was taken from the 'abbrevated' RDF/XML syntax in the 
RDF Model & Syntax Rec, which has been superceded by RDF Syntax 
(Revised) which does not use this term)

Most of the pictures lost their right edge when I printed them off.


A-Box and T-Box should have different namespaces.
=================================================

The typical user would be importing the ontology and creating their 
instances in their own namespace (if the TF wanted to address this then 
also the individual names for the participants in the relationship are 
impacted)
i.e.
in

http://www.w3.org/2001/sw/BestPractices/OEP/n-aryRelations-20040623/diagnosis.owl

http://protege.stanford.edu/diagnosis#diagnosis_probability

is in the ontology (T Box),
and
http://protege.stanford.edu/diagnosis#Christine

is in the A Box. In practice, we would expect these to be in different 
spaces. I suggest that the ontology should be in

http://www.w3.org/2001/sw/BestPractices/n-aryRelations/diagnosis#

and the A Box in

http://www.w3.org/2001/sw/BestPractices/n-aryRelations/diagnosis-example#

etc. for the other examples.

Technical issues with example files
===================================
Looking at the diagnosis files ...

N3

1) the comment concerning the base URL at the top of file has no effect, 
and is just a comment

2)

(Sorry I need to go now, I will complete the technical review of the 
example files later)

Jeremy
Received on Tuesday, 6 July 2004 06:08:42 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:17:12 GMT