Recommended reading: feedback on SKOS from Doug Tudhope and Ceri Binding from University of Glamorgan

Forwarding this to the list.  

-----Original Message-----
From: Douglas Tudhope [mailto:dstudhope@glam.ac.uk] 
Sent: 05 February 2004 16:04
To: Miles, AJ (Alistair) 
Cc: cbinding@glam.ac.uk
Subject: feedback on RDF schemas for thesauri and simple KOS *pre-release*


Alistair

SKOS-CORE looks  a very promising start.
As promised (belatedly - sorry) some initial comments on these from our
point of view (bearing in mind that we don't have much experience in RDF).

1. As you know, we support the idea of allowing for more precise KOS
representations. We should also try to maintain compatability with
traditional standards. One rationale for the current standard's set of
thesaurus relationships is that they are at quite a cost/effective level of
generality for many applications, allowing for some user/indexer variation
in concept useage and relevance judgements. The proposed scheme does seem to
allow for both 'traditional' KOS and more precise, formal constructions
which is good.

2.  It is important to have some notion of facets but we don't think that
current version quite captures it. The scheme correctly takes 'facets' to
represent fundamental categories in the sense of Ranganathan, the CRG, BSI
standard etc (as opposed to subfacet indicators). Yes, each concept is a
member of one and only one facet. But in this sense, I'm not sure it's
useful to simply 'treat facets as concepts'? Eg
>>>
- <rdf:Property rdf:ID="inFacet">
  <rdfs:label>member-of-facet</rdfs:label>
  <rdfs:subPropertyOf rdf:resource="#broader" />
  <rdfs:range rdf:resource="#Facet" />
  <rdfs:comment> This property indicates that a concept is a member of a
facet. A concept may have only one inFacet property. This property is a
sub-property of the 'broader' property. Thus faceted conceptual structures
may be reduced to simple hierarchical displays by applications that do not
comprehend facets.</rdfs:comment>
  </rdf:Property>
>>>
a) "This property is a sub-property of the 'broader' property. "  -- In
particular, we're not convinced about making this relationship a type of the
broader relationship. I'd suggest that's not really the semantics?
Conceivably subclass/superclass or set membership might be better solutions?

b) In any case, we don't quite see how this would realise the degenerate
case you describe for non-faceted schemes. Each concept is immediately
related by the BT-subtype relationship to its Facet. How does that help you
with the degenerate case, where presumably you want the Top of Hierarchy to
stand in as a facet? A subclass would do just as well?

c) The degenerate case is less important than being able to facilitate more
advanced reasoning with faceted schemes. Thus we might wish to use OWL or
another language to express facet synthesis rules, or relate thesaurus
facets to a higher level ontology. There needs to be a sufficiently clear
distinction between facets and member concepts - use of the broader
relationship concerns us here.


3. Broader/Narrow - We assume that the OWL representation would formally
express the Inverse relationship? However, in the RDF there is nothing to
capture the BT/NT connection apart from the comment. Is it useful to make
the RDF as self sufficient as possible? Eg would it be useful to introduce a
new type of Semantic Relationship called HierarchicalRelationship with BT
and NT underneath? We can see this may have drawback of creating additional
complexity but suggest as a consideration.


4. Not everyone considers Related (RTs) to be necessarily symmetric (eg the
AAT does not). Could 'symmetric' not be an optional property of the
relationship?

5. Is there any possibility of defining at least one subtype of Related? Eg
a Partitive (see below)?

6. Good to have subtypes of the hierarchical relationships but note that
broader/narrowerPartitive is often restricted to members of the same
hierarchy (see Aitchison&Gilchrist). In other cases, a Related relationship
type is recommended.


7. There is little notion here of the Entry Vocabulary, and the various
relationships between concepts and terms. Was there a reason for this? It's
a very important aspect of a thesaurus and may(?) be a critical issue for
gaining acceptance of  a standard in some traditional thesaurus circles. It
take it the rationale is http://esw.w3.org/topic/RdfThesaurus - "Suggested
solution 2:: We don't bother with them. Instead we offer the recommendation
that all acronyms be included as possible labels for a concept. Plural forms
probably don't need be included as modern stemming algorithms can identify
the root of the term. "

a) I see the general point and agree that the longterm solution is making
connections with standards in the Linguistic community. However we would
tend to argue that relying only on labels results in an impoverished model
of a thesaurus? Essentially a thesaurus contains a pragmatic domain-specific
lexicon in the entry vocab, equivalence relationships and scope notes. It's
one of the reasons why the thesaurus has been such a useful tool over many
years and arguably a weakness of some purely concept-based 'ontological'
efforts. Could we bring in some version of the Equivalence relationship to
SKOS, or alternatively have more properties regarding terms?

b) For example, there are various subtypes of equivalence corresponding to
parts-of-speech relationships, US/UKalts, types of synonyms, antonyms (even)
and these might be distinguished in some future super-KOS systems. Replacing
the Equivalnce relationship with a simple 'bag of labels' would lose that
possibility.

c) Also - in some cases a term will be considered Equivalent to more than
one concept (perhaps with different degrees of confidence). Again that
becomes less clearly stated.



8. Anyway, be interested to know your thoughts on all this - it's great that
someone is proposing possible standards and trying to reach concensus. The
general thrust of the SKOS RDF schema is great. What are your plans for
progressing it?

One way of possibly progressing/discussing some of this effort might be in
(or associated with) an NKOS workshop at ECDL'04 in Bath, this September.
Marianne Nielsen is aiming to propose a workshop on user-centred issues and
this might be a second (or parallel mini-meeting) theme?

Hope to get a chance to talk at JISC workshop in London at some point if you
are attending for SWAD-Europe demo?

regards

Doug, Ceri


Douglas Tudhope
Reader, School of Computing
University of Glamorgan
Pontypridd CF37 1DL
Wales, UK

Tel  +44 (0) 1443-482271
Fax  +44 (0) 1443-482715
dstudhope@glam.ac.uk http://www.comp.glam.ac.uk/pages/staff/dstudhope
Editor : The New Review of Hypermedia and Multimedia


----- Original Message -----
From: "Miles, AJ (Alistair)" <A.J.Miles@RL.AC.UK>
To: <NKOS@dli2.nsf.gov>
Sent: Tuesday, November 25, 2003 1:41 PM
Subject: RDF schemas for thesauri and simple KOS *pre-release*


> Dear all,
>
> I offer these schemas as a pre-release, to get some initial feedback
and
> response on their design.
>
> SKOS-Core <http://www.w3c.rl.ac.uk/2003/11/21-skos-core> (RDF schema
for
> encoding thesauri and other simple knowledge organisation systems e.g. 
> taxonomies and classification schemes.)
>
> SKOS-Mapping <http://www.w3c.rl.ac.uk/2003/11/21-skos-mapping> (RDF 
> schema for expressing mappings between concepts from different 
> thesauri.)
>
> This work is ongoing in the context of the SWAD-Europe project [1]
[2].
>
> Yours,
>
> Alistair Miles.
>
> [1] SWAD-Europe Thesaurus Activity 
> <http://www.w3c.rl.ac.uk/SWAD/thesaurus.html>
> [2] Semantic Web Advanced Development for Europe project 
> <http://www.w3.org/2001/sw/Europe/>
>
>
> CCLRC - Rutherford Appleton Laboratory
> Building R1 Room 1.60
> Fermi Avenue
> Chilton
> Didcot
> Oxfordshire OX11 0QX
> United Kingdom
>
> Email:        a.j.miles@rl.ac.uk
> Telephone: +44 (0)1235 445440
>

Received on Tuesday, 24 February 2004 08:22:01 UTC