W3C home > Mailing lists > Public > public-owl-dev@w3.org > October to December 2006

RE: OWL "Sydney Syntax", structured english

From: Pat Hayes <phayes@ihmc.us>
Date: Wed, 29 Nov 2006 15:21:28 -0600
Message-Id: <p0623090bc193a680f548@[10.100.0.128]>
To: "John McClure" <jmcclure@hypergrove.com>
Cc: "Kaarel Kaljurand" <kaljurand@gmail.com>, "Dan Connolly" <connolly@w3.org>, "Anne Cregan" <annec@cse.unsw.edu.au>, <public-owl-dev@w3.org>

>Pat,
>It would truly be a shame to ignore the usefulness of verbs 
>altogether. Not only
>does Dan know Pat, but Dan once knew John, or Pat should know John, or Dan may
>know Pat's brother, and so on.... currently RDF semantics only focuses on
>present transitives apparently without any idea how to evolve to a more
>"structured english syntax" that incorporates matters of time and conditions.

No, wait. The RDF semantics does not focus on present transitives. It 
says nothing whatever about time, tense or transitives. It would be 
very dangerous indeed to read RDF as being all written in the English 
present tense. It might be a good idea to use the OWL-TIME ontology 
as a test case, in fact, for any of these proposals.

>
>This is exactly the problem that I have been addressing the last few 
>years.  In
>Version 1 of the Legal-RDF ontology, I started with striped syntax, 
>yielding for
>example
>   <Person rdf:ID='Scarecrow'>
>      <has>
>         <Brain/>
>      </has>
>   </Person>
>defining Person and Brain as classes, and defining has, had, willHave, etc as
>'properties'. But now, I am working on Version 2, with infinitely better
>results, defining Brain as a TRUE property whose range is BrainMatter:

My goodness. I am at a loss to understand exactly what that is 
supposed to mean, I have to confess. It sounds like a mereological 
notion (?) What are the members of the class BrainMatter supposed to 
actually be? Lumps of neural tissue?

>   <Person rdf:ID='Scarecrow'>
>      <Brain lgl:verb='has'>
>          <BrainMatter/>
>      </Brain>
>   </Person>
>
>I provide background about all this at
>http://aufderheide.info/lexmlwiki/index.php?title=Legal-RDF_Ontologies showing
>also how I am tying RDF into ECMA languages (basically, just as Dan 
>suggests in
>another note).  Anyway, my point is that verbs are powerful linguistically for
>solid semantic reasons -- these must be incorporated into reasoning tools but
>most definitely NOT via the names of properties of classes of individuals...

Well, you keep saying this, but I havnt seen a single reason offered 
for WHY you are saying it.

>
>Simply, I can't accept the assertion that the property set of a 
>Person includes
>a "hasBrain".

I really don't know what you mean by 'the property set of a Person' 
here. Properties in RDF and OWL are simply binary relations between 
things. I have no trouble with the property of being a thing that has 
a brain: it could be defined as the restriction on a property 
indicating a type of nervous system, for example.

>Thanks,
>John
>
>PS Most OWL properties are verb+preposition properties.  But it also defines
>verb+noun properties -- eg hasValue.

Well, actually hasValue isn't a property in OWL, it is part of the 
syntax defining a restriction on properties. Perhaps we are using the 
word 'property' at cross purposes?

>  And it defines verb-only properties -- eg
>imports.  And it defines noun-only properties -- eg cardinality.  In other
>words, it seems rather arbitrary what is best defined as a property 
>and what is
>best defined as a class.

In RDF/OWL, classes are unary relations and properties are binary 
relations. There is some arbitrariness here, as you indicate.

I have no idea what you mean by noun-only and verb-only.

>  RDF defines noun-only properties incidentally: type,
>subject, predicate, object and domain are all nouns.  RDFS actually got the
>confusion started -- subClassOf rather than superclass is a good example.

The convention illustrated here did not start with RDFS. It goes 
back, and has been widely used, for at least two decades before that, 
including the capitalization convention. These are very old debates.

>rdfs:label is a noun only.... all Dublin Core properties are nouns....

Right, all these are plausibly regarded as 'roles', conventionally 
rendered in English using nouns, often with 'of'. But not all 
relationships are roles.

>FOAF,
>constantly made an exemplar by W3, is a bit of a mess in this regard IMHO.

IMO, any attempt to avoid 'mess' in your sense will be doomed to 
failure, and is best not even attempted. It has been tried before, 
and it has always failed. And in any case there is absolutely no need 
to even attempt it. Nothing turns on the noun/verb distinction.

>
>PPS Negation is straight-forward under my approach
>
>   <Person rdf:ID='Scarecrow'>
>      <Brain lgl:verb='hasNot'>
>          <BrainMatter/>
>      </Brain>
>   </Person>

Something has to know that hasNot... means not (has ...)

Pat

-- 
---------------------------------------------------------------------
IHMC		(850)434 8903 or (650)494 3973   home
40 South Alcaniz St.	(850)202 4416   office
Pensacola			(850)202 4440   fax
FL 32502			(850)291 0667    cell
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes
Received on Wednesday, 29 November 2006 21:21:39 GMT

This archive was generated by hypermail 2.3.1 : Wednesday, 27 March 2013 09:32:53 GMT