Re: [Proposal][SKOS-Core] skos:denotes

In message <20040929184124.GW2012@homer.w3.org> on Wed, 29 Sep 2004, Dan 
Brickley <danbri@w3.org> wrote
>
>Hi Stella,
>
>* Stella Dextre Clarke <sdclarke@lukehouse.demon.co.uk> [2004-09-29 18:20+0100]
>>
>> Sorry, but I am completely lost in this conversation.
. . .
>> I think I am saying you are  asking for the moon, and any 
>>approximation to the moon could prove  unsatisfactory.
>
>It's a tricky idea to explain, but I've seen enough RDF applications
>that describe people and other entities directly that I know it's not a
>moon-on-stick. So I think the problem is with communicating the basic idea

I think that my position is much the same as Stella's - we both come 
from a background in traditional thesauri, and are having difficulty in 
coming to terms with the complexity that is being introduced in encoding 
schemes to make machines understand them.

Can  I crave your patience if I try to sort this out in my head from 
first principles?

It seems to me that we have four things to build with:

1. concepts
2. attributes of concepts
3. relationships between concepts
4. attributes of relationships

Attributes can be decomposed into further concepts and relationships, so 
that instead of saying that a particular person (an instance of the 
concept of "persons") has IBM as an employer, a height of 1.8m and an 
age of 52 years, we can represent these as relationships between the 
person and the concepts of "organisation", "height" and "age", 
respectively.

The relationship with "organisation" is "employed by" and may have a 
date range as an attribute, while the relationships with height and age 
each have to contain a numeric value and a unit as attributes.

>For cases like people, places, organisations, this is quite important, 
>since non-SKOS RDF can directly model some very detailed 
>characteristics of these entities. We can use non-SKOS RDF to describe 
>a Person's age, height, employer, etc.; or a place's latitute and 
>longitude, or an Organization's certifications. That's the sort of task 
>RDF was built for.

So RDF to some extent keeps attributes of concepts undecomposed. A 
thesaurus or "authority file" can also do this, though it may express 
attributes as notes rather than in a more structured fashion. In a 
thesaurus the types of relationship between concepts that can be 
expressed are limited, and relationships do not generally have 
attributes.

>SKOS uses W3C's Resource Description Framework (RDF) to describe
>thesauri (and similar things).

So does RDF provide a mechanism for recording all of the items 1 to 4 I 
have listed above?  If so, I don't see why RDF has difficulty in 
encoding all the concepts of a thesaurus. You say

>RDF's perspective on the world doesn't work so well "out of the box"
>when trying to model things like fluids / mass nouns, or processes. So
>there are often enough concepts in a thesaurus that don't have an
>obvious correlation in non-SKOS RDF descriptions.

Is this just saying that some of these more abstract or general concepts 
cannot easily be encoded in an RDF structure, or just that they may not 
often occur in particular existing ontologies that have been created 
using that structure?

I think my difficulty is in grasping the way that RDF defines concepts. 
To me a concept is "a unit of thought" (the definition we have put in 
the draft revision of the British Standard for thesaurus construction). 
The scope of meaning of a concept within a thesaurus is given by its 
relationships and, preferably, by an explicit scope note.

Al contrasts this with

>The 'ontology style'

which

>describes things in terms of their properties (i.e. with *logical* 
>constructs).

I really don't think that this is significantly different. A good scope 
note in a thesaurus defines a concept in terms of the broader category 
to which it belongs together with the properties that distinguish that 
concept from other concepts within the broader category.

If, therefore, the idea of a "concept" is essentially the same in both 
types of scheme, I don't know why we need a special type of relationship 
to express exact or approximate equivalence when combining schemes or 
mapping between them, other than the relationship that would be used 
when combining or mapping schemes of the same type.

>In other words, when people ask us "Should I be creating an Ontology or
>a Thesaurus?", we want to figure out what to tell them (beyond "yes!")
>about the diffent approaches possible using SKOS, RDF, OWL etc. And
>hopefully have some conventions for connecting up information created in
>either style.

My hope is that rather than needing conventions for connecting up 
conflicting styles we can have an umbrella scheme where there is only 
one definition of each element (such as "concept"), and that we should 
have the components available to allow us to express any additional 
aspects, such as many types of relationship and attributes of concepts 
and relationships. A thesaurus structure such as SKOS would then just 
need a subset of these elements.

(I have omitted discussion of facets and arrays, so as not to complicate 
the issue for the moment.)

Leonard

-- 
Willpower Information       (Partners: Dr Leonard D Will, Sheena E Will)
Information Management Consultants              Tel: +44 (0)20 8372 0092
27 Calshot Way, Enfield, Middlesex EN2 7BQ, UK. Fax: +44 (0)870 051 7276
L.Will@Willpowerinfo.co.uk               Sheena.Will@Willpowerinfo.co.uk
---------------- <URL:http://www.willpowerinfo.co.uk/> -----------------

Received on Wednesday, 29 September 2004 22:10:09 UTC