W3C home > Mailing lists > Public > public-swbp-wg@w3.org > May 2004

Re: [WNET, PORT, OEP] Synset's and Classes - dumb question

From: Dan Brickley <danbri@w3.org>
Date: Wed, 05 May 2004 19:38:00 +0100
Message-ID: <40993488.7090102@w3.org>
To: "Miles, AJ (Alistair) " <A.J.Miles@rl.ac.uk>
Cc: "'public-swbp-wg@w3.org'" <public-swbp-wg@w3.org>, "'a.gangemi@istc.cnr.it'" <a.gangemi@istc.cnr.it>

Miles, AJ (Alistair) wrote:

> Hi Aldo, all,
> 
> 
>>>Is there a consensus view on the relationship between a 
>>
>>wordnet synset and
>>
>>>the class the synonyms names, i.e. is the synset containing 
>>
>>the word 'dog'
>>
>>>necessarily owl:sameAs the class of dogs?
>>
>>owl:sameAs applies to owl:Individuals, so you are asking a meta-level 
>>question :)
>>
>>OK, my position is that - provided that we want to transform a 
>>wordnet into a formal ontology - the semantic interpretation of 
>>"synset" is that of an equivalence class of words/terms according to 
>>a common intended meaning. Since "having a unique intended meaning" 
>>is also applicable to classes, the *default* mapping of synsets is to 
>>owl:Class.
>>On the other hand, not only classes have a unique intended meaning, 
>>but also individuals, and as a matter of fact, many synsets refer to 
>>individuals like "Italy" or "Cicero". That's why "synset" hasn't a 
>>precise mapping to formal ontologies. Then, your dog example is 
>>correct, but not "necessarily".
> 
> 
> The skos:Concept class [1] is there to type something that has a unique
> intended meaning, but is not necessarily a class or an individual.  Do you
> think it is appropriate to map synsets onto skos:Concepts?
> 
> Dan Brickley and I were discussing this sort of thing recently.  We wanted
> to be able to express the relationship between for example the concept that
> I label with the word 'dog', and the class of dogs.  There are some
> philosophical and (far more tractable) practical issues that arise if you
> alow these two nodes to be merged (i.e. if you use owl:sameAs).  We arrived
> at a tentative suggestion to introduce a predicate into the SKOS vocab,
> called skos:conceptualises (or something along the same lines).  So e.g.
> 
> [	a	skos:Concept;
> 	skos:prefLabel	'dogs';
> 	skos:altLabel	'mutts';]
> skos:conceptualises
> [	a	owl:Class;
> 	rdfs:label		'The class of all dogs';]
> .
> 
> I keep trying to write this idea up, but the emails get so long think no one
> will ever get to the end of them!  Anyway, this is a strawman, what do you
> think?

I think something like this is both needed and useful. Aldo very nicely 
outlined how ontological, thesaurus and wordnet/dictionary perspectives 
relate. Wordnet's noun term synsets are really interesting here, as they 
sit right in the middle, since they imposed a discipline that "___ is a 
kind of ___" must relate each hypernym pair (ignoring for now the 
individuals like "Paris" which mess this up).

There are lots of reasons for trying to take an RDF/OWL perspective 
wherever possible, but there are also plenty of cases where a 
nevertheless-useful structure (eg. thesauri, also bookmark hierarchies 
and other scruffy organisational systems) is better represented in 
something like SKOS. Lexically driven electronic dictionaries (wordnet, 
eurowordnet, the EDR in Japan, etc) have relationships that make sense 
between words, concepts and suchlike, but don't really make sense in the 
class and property-centric world of ontologies.

I'm not really sure what to make of the phrase "unique intended 
meaning", but... to my mind the notion of "concept" in SKOS is loose 
enough to cover dictionary-style content. I think it might be worth 
experimenting with expressing Wordnet and similar with extensions 
applied to a basic structure drawn from SKOS, with synsets being treated 
as skos concepts.

The immediate need, imho, is to be able to indicate when a SKOS-style 
structure has nodes which stand for things that vanilla RDF/OWL sees as 
individuals or classes. For example, a concept graph that had a node for 
"Vegetarian restaurants in Bristol" could be associated with a class; 
one that had a node for "Dan's Diner" could be associated with an 
individual of type xyz:Restaurant that had a homepage, location, owner 
etc., all describable in  RDF.

In other words, we need both styles of representation, and a way of 
indicating some basic relationships between the two perspectives.

cheers,

Dan

> 
> Alistair.	     
> 
> [1] http://www.w3.org/2001/sw/Europe/reports/thes/1.0/guide/
Received on Wednesday, 5 May 2004 14:38:13 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:09:38 UTC