W3C home > Mailing lists > Public > public-swbp-wg@w3.org > November 2005

[PORT] comments on SKOS

From: Jeremy Carroll <jjc@hpl.hp.com>
Date: Tue, 15 Nov 2005 16:28:41 +0000
Message-ID: <437A0CB9.4010501@hpl.hp.com>
To: public-swbp-wg@w3.org


reviewed documents

[[

PORT TF

     Documents:

     SKOS Core Guide
     http://www.w3.org/2004/02/skos/core/guide/2005-10-06/

     SKOS Core Vocabulary Specification
     http://www.w3.org/2004/02/skos/core/spec/2005-10-06/

     Discussion:

     Message Miles;
     http://lists.w3.org/Archives/Public/public-swbp-wg/2005Oct/0195.html

     SKOS change proposals:
     http://www.w3.org/2004/02/skos/core/proposals#coordination-8
]]


SKOS Core Guide

1) suggest only having XHTML and CSS valid buttons to bottom of
document, for greater consistency with other W3C docs. (although in
editors draft useful to have them at top)

2) a number of things printed badly, e.g. many of the pictures did not
fit on the page, also the extract from UKAT did not wrap

In "Introduction", para starting "SKOS Core is an application ..."
suggest reference to RDF Primer would be appropriate for intended
audience, rather than RDF Concepts, RDF Semantics, RDF Syntax

3) Also in Introduction, suggest other prerequisites should be
discussed, not just SW knowledge. e.g. some background in taxonomies or
whatever

4) subsection entitled "Examples", suggest that ex:aResource etc. should
be more meaningful names

5) Reference for URI is RFC 3986 (or maybe IRI 3987)

6) Suggest new section between "Labelling Properties" and "Multilingual
Labelling" introducing mono-lingual labelling, noting that the
assumption that the reader knows that the labels are in English is in
general false on the Web

7) Suggest that all literals with natural language text other than in
the first section on "Labelling Properties" should have a language tag
(e.g. @en) [This is a big change in terms of number of changes]

8) question on dcterms:RFC1766, that RFC was updated a long time ago to
RFC 3066, and that in turn is in the process of being updated ... does
dcterms:RFC1766 update itself or is it specifically linked to that version

9) The sentence "The property skos:related is a symmetric property."
implies that the example abaove has a redundant triple. This perhaps
should be clarified, e.g. adding to the above sentence ", and so one of
the triples can be omitted."

10) Suggest not using the word "recommended" since this is not a
recommendation.

11) Semantics of skos:hasTopConcept does have semantic import.
e.g. the following seems to be an inconsistent SKOS document:

_:a skos:hasTopConcept eg:top .
eg:top skos:narrower eg:notTop .

or maybe it implies that

eg:notTop skos:narrower eg:top .

12) Need to clarify whether eg:narrower is reflexive or irreflexive

13) Most readers will be making a unique names assumption, but RDF in
general does not.
The labelling may allow the ability to force unique naming by some
semantic constraint, e.g. skos:prefLabel as an
InverseFunctionalProperty, (although that's outside OWL DL)

This relates to the above example, in that if skos:narrower is
reflexive, and skos:prefLabel is not an InverseFunctionalProperty or
similarly constrained then

_:a skos:hasTopConcept eg:top .
eg:top skos:narrower eg:notTop .
eg:top skos:prefLabel "Top" .
eg:notTop skos:prefLabel "Not Top" .

entails

eg:top owl:sameAs eg:notTop .
eg:notTop skos:prefLabel "Top" .
eg:top skos:prefLabel "Not Top" .

If there is an IFP like constraint, or if skos:narrower is irreflexive,
then this is inconsistent.


14) Collectable Properties stuff
a question ....

If we divide the concept of human up into two concepts adults and
children, and also up into two concepts of males and females, then adult
is narrower than human

<skos:Collection>
    <skos:member rdf:resource="&eg;male"/>
    <skos:member rdf:resource="&eg;female"/>
</skos:Collection>

<skos:Collection>
    <skos:member rdf:resource="&eg;adult"/>
    <skos:member rdf:resource="&eg;child"/>
</skos:Collection>

and

<skos:Concept rdf:about="&eg;male" />
<skos:Concept rdf:about="&eg;female" />
<skos:Concept rdf:about="&eg;adult" />
<skos:Concept rdf:about="&eg;child" />

<skos:Concept rdf:about="&eg;human" />

What are the skos:broader and skos:narrower relationships between these
5 concepts and 2 collections?


15) The para at start of subsection "Concetps in Multiple Schemas" is
incorrect. Of course RDF statements change the meaning of the things
that they are about; and not only the informal descriptions. The example
in point 13 explores one case, the collectable properties rule is
another example.


16) It would be nice if skos:prefLabel had a uniqueness property, e.g.
maybe have skos:SingleLabelScheme rdfs:subClassOf skos:ConceptScheme,
for which there is a unique naming assumption on the labels with any
given language tag.



On

     SKOS Core Vocabulary Specification
     http://www.w3.org/2004/02/skos/core/spec/2005-10-06/

17)Change Policy Statement

Has this policy been followed so far?
e.g. publish every two months?

The policy should be reviewed to ensure that it is an honest statement
of aspiration as possible.



Comment on proposed section Example:Weighted Semantic Relationships

I don't like this at all.

RDF reification vocabulary is normative.
It's normative semantics is very weak.

The final comment suggesting the reification of a statement entails the 
triple is incorrect, and poorly phrased.


Comment on SOlution for olwImport-7

I very much like additional notes 3 and 4

Note 3: factoring out the english lexical resources:
    - yes english is the default language in W3C
    - but also yes the english labels should be accessible using the 
same mechanisms as any other supported language. This will allow tools 
to not have to special case for english.

Note 4: factoring out is good; factoring out the housekeeping is likely 
to be particularly beneficial.



Jeremy
Received on Tuesday, 15 November 2005 16:29:08 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:17:19 GMT