W3C home > Mailing lists > Public > public-esw-thes@w3.org > July 2008

RE: aboutness and broader

From: Alistair Miles <alistair.miles@zoo.ox.ac.uk>
Date: Fri, 25 Jul 2008 11:34:35 +0100
To: "'Jonathan Chetwynd'" <j.chetwynd@btinternet.com>
Cc: <public-esw-thes@w3.org>
Message-ID: <004301c8ee42$08f53900$1adfab00$@miles@zoo.ox.ac.uk>

Hi Jonathan,

Apologies, I could have avoided using "necessarily" altogether. I meant
something informal. I was just sharing a thought about how a commonly used
pattern for using structured vocabularies in information retrieval systems
could be implemented with RDF/OWL.

The pattern is this: 

Given a structured vocabulary X, Y, Z, arranged in a broader/narrower
hierarchy where X is broader than Y and Y is broader than Z.

Given a set of documents indexed with this vocabulary. 

The user queries for X, and the retrieval system returns documents indexed
with X plus documents indexed with Y plus documents indexed with Z.

E.g. this is how most blogging systems handle search using a hierarchical
categorisation scheme.

I was just pointing out that, one way to implement this information
retrieval functionality using RDF/OWL is this:

Given an RDF property to represent the indexing link between a document and
a vocabulary concept e.g. dc:subject

Given a property to represent the broader hierarchical link between
vocabulary concepts e.g. skos:broader

Declare a property chain axiom e.g. the property chain (dc:subject,
skos:broader) is a sub-property of dc:subject. 

...This property chain axiom then provides the basis for inferring an
"expanded" index, against which you can evaluate queries. So nothing deeply
philosophical, just a practical way of using an inference pattern to
implement a common information retrieval functionality. There are of course
a number of other ways to implement the same pattern, all of which are
possible with the current SKOS data model ... someone should probably write
a note about this sometime (hint to self :)

Btw property chains are a proposed feature in OWL 2, and a very useful one
IMO. The notion of property chains is currently used in the SKOS Reference
to define the dependencies between the SKOS XL extension and vanilla SKOS,
which then provides the basis for "dumbing down" SKOS+XL to vanilla SKOS.

Hope I haven't added to the confusion.

Alistair.

--
Alistair Miles
Senior Computing Officer
Image Bioinformatics Research Group
Department of Zoology
The Tinbergen Building
University of Oxford
South Parks Road
Oxford
OX1 3PS
United Kingdom
Web: http://purl.org/net/aliman
Email: alistair.miles@zoo.ox.ac.uk
Tel: +44 (0)1865 281993

> -----Original Message-----
> From: public-esw-thes-request@w3.org [mailto:public-esw-thes-
> request@w3.org] On Behalf Of Jonathan Chetwynd
> Sent: 04 June 2008 10:31
> To: Alistair Miles
> Cc: public-esw-thes@w3.org
> Subject: Re: aboutness and broader
> 
> Alistair,
> 
> as always with issues around logic, the reasoning isn't always apparent
> to the outsider...
> 
> I'm not clear about your use of "necessarily", please can you expand
> briefly?
> 
> ie what does "mammals have breasts" tell us about animals?
> or "carrots are orange" tell us about vegetables?
> other than some B can be D? but not necessarily?
> 
> and how does that relate to "animals are motile" not telling us about
> mammals?
> ie all mammals are motile? is necessarily? though...
> 
> cheers
> 
> ~:"
> 
> please excuse my ignorance in this field
> 
> my logo <http://www.openicon.org>
> 
> Jonathan Chetwynd
> 
> j.chetwynd@btinternet.com
> http://www.openicon.org/
> 
> +44 (0) 20 7978 1764
> 
> 
> On 4 Jun 2008, at 10:14, Alistair Miles wrote:
> 
> 
> 
> 	Hi all,
> 
> 	I thought I'd write a quick comment on how "broader" interacts
> with the
> 	notion of "aboutness", and how this could be captured formally.
> 
> 	A common view is that, if some concept B is broader than A, then
> if a
> 	document D is "about" A, D is necessarily also "about" B.
> 
> 	(E.g. all documents "about" mammals are necessarily also "about"
> animals.)
> 
> 	To capture this view formally using RDF and OWL, we first need an
> RDF
> 	property to represent our notion of "aboutness". For the sake of
> 	illustration, let's use dc:subject.
> 
> 	We then express a property chain axiom. We say that the property
> chain
> 	(dc:subject, skos:broader) is a sub-property of dc:subject.
> 
> 	Now, given this property chain axiom, the graph
> 
> 	<D> dc:subject <A>.
> 	<A> skos:broader <B>.
> 
> 	entails
> 
> 	<D> dc:subject <B>.
> 
> 	Notice that the property chain axiom causes the dc:subject link
> to
> 	"propagate" up the concept hierarchy until it reaches the top.
> I.e. the
> 	graph
> 
> 	<D> dc:subject <A>.
> 	<A> skos:broader <B>.
> 	<B> skos:broader <C>.
> 
> 	entails
> 
> 	<D> dc:subject <B>, <C>.
> 
> 	Notice also that this behaviour only depends on the property
> chain axiom. It
> 	does not require that skos:broader be transitive.
> 
> 	Cheers,
> 
> 	Alistair.
> 
> 	--
> 	Alistair Miles
> 	Senior Computing Officer
> 	Image Bioinformatics Research Group
> 	Department of Zoology
> 	The Tinbergen Building
> 	University of Oxford
> 	South Parks Road
> 	Oxford
> 	OX1 3PS
> 	United Kingdom
> 	Web: http://purl.org/net/aliman
> 	Email: alistair.miles@zoo.ox.ac.uk
> 	Tel: +44 (0)1865 281993
> 
> 
> 
> 
> 
> 
Received on Friday, 25 July 2008 10:35:15 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:39:00 GMT