W3C home > Mailing lists > Public > public-esw-thes@w3.org > March 2004

Modelling 'term-to-term' relationships in SKOS

From: Miles, AJ (Alistair) <A.J.Miles@rl.ac.uk>
Date: Wed, 3 Mar 2004 09:01:45 -0000
Message-ID: <350DC7048372D31197F200902773DF4C047485F4@exchange11.rl.ac.uk>
To: Douglas Tudhope <dstudhope@glam.ac.uk>
Cc: public-esw-thes@w3.org

I designed SKOS to be strictly concept-oriented.  In this view there are
only 'concepts' and 'labels'.


In the concept-oriented view, 'antonymy' would be represented as a
relationship between two concepts.  So in an imaginary thesaurus, concept A
(label 'Black') is opposite of concept B (label 'White').  As an extension
of SKOS, the relationship 'is opposite of' would be modelled as a
sub-property of skos:SemanticRelation.

US/UK Alts:

If distinguishing between US/UK alts is important, they could both be
represented as labels with language tags, treating US-English and UK-English
as separate languages.  E.g.

	<skos:prefLabel xml:lang="en-US">Color</skos:prefLabel>
	<skos:prefLabel xml:lang="en-GB">Colour</skos:prefLabel>

[RDF-folks: can we do this ... have I got the codes right for the language
tags of the literals?]

I suspect most of the time it will be sufficient to include US spelling alts
among the altLabels (if it is an english thesaurus), without needing to add
language tags.  This would be enough to ensure concepts are picked up by
searching yanks.

Types of synonyms and parts of speech:

Doug: can you elabourate on what these are for us?

Equivalence between terms:

In the concept-oriented view, a label is simply a string.  Therefore, there
is no notion of 'equivalence' between labels.

A label may be used as a label for more than one concept.  However, a label
is always considered to be empty of meaning.  Therefore, there is no notion
of 'equivalence' between a label and a concept.  

SKOS supports a notion of 'equivalence' between concepts.  This may either
be modelled as a sub-property of skos:SemanticRelation (i.e. an
intra-thesaurus relation) or a sub-property of skos-map:SemanticMapping
(i.e. an inter-thesaurus relation).  

E.g. SKOS-Mapping contains the property skos-map:exactMapping, which could
be used to express an 'exact equivalence' relationship between two concepts.

E.g. SKOS-Mapping contains two properties skos-map:MajorMapping and
skos-map:MinorMapping, which could be used to express greater or lesser
degrees of 'equivalence' between concepts.

Final word:

This is how I think of it.  To switch from a 'term-oriented' view to a
'concept-oriented' view, instead of 'term' think 'concept+label'.

What does everyone think?


Doug wrote:
> However there are some other possibilities that may be more 
> difficult to
> implement without explicit equivalence relationships.
> - various subtypes of equivalence corresponding to
> parts-of-speech relationships, US/UKalts, types of synonyms, 
> antonyms (even)
> - in some cases a term can be considered Equivalent to more than
> one concept (perhaps with different degrees of confidence).
> I guess these are more future application possibilities, might
> be considered as less 'core' and we may be in danger of 
> over-elaborating.
> However if we go on to extend the current core thesaurus
> relationships by specialisation then we might also want to distinguish
> subtypes of equivalence. This might be easier if there was an explicit
> equivalence relationship.
> Doug

Alistair Miles
Research Associate
CCLRC - Rutherford Appleton Laboratory
Building R1 Room 1.60
Fermi Avenue
Oxfordshire OX11 0QX
United Kingdom
Email:        a.j.miles@rl.ac.uk
Tel: +44 (0)1235 445440 
Received on Wednesday, 3 March 2004 04:02:19 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 19:45:10 UTC