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 13:58:50 -0600
Message-Id: <p0623090ac1938f688ba7@[]>
To: Dan Connolly <connolly@w3.org>
Cc: John McClure <jmcclure@hypergrove.com>, Kaarel Kaljurand <kaljurand@gmail.com>, Anne Cregan <annec@cse.unsw.edu.au>, public-owl-dev@w3.org

>On Wed, 2006-11-29 at 12:45 -0600, Pat Hayes wrote:
>>  >  >-----Original Message-----
>>  >>From: public-owl-dev-request@w3.org
>>  >>[mailto:public-owl-dev-request@w3.org]On Behalf Of Kaarel Kaljurand
>>  >>Sent: Wednesday, November 29, 2006 2:58 AM
>>  >>To: Dan Connolly
>>  >>Cc: Anne Cregan; public-owl-dev@w3.org
>>  >>Subject: Re: OWL "Sydney Syntax", structured english
>>  >>
>>  ><snip/>
>>  >>No, the expectation is that every propertyname is a transitive English
>>  >>verb, rather than a noun. There are studies that seem to indicate that
>>  >>this is the case in reality, see e.g.:
>>  >  > http://www.csd.abdn.ac.uk/~cmellish/papers/kbs05.pdf
>>  ><snip/>
>>  >
>>  >Hmm.
>>  >
>>  >While I personally and professionally support the aims of a 
>>structured english
>>  >syntax, I am wondering when the RDF community will align their 
>>practices with
>>  >those of XML standards communities, which seem 100% to eschew the 
>>use of verbs
>>  >as property-names.
>>  WHY?? Properties relate things together. One makes an assertion,
>>  typically, by connecting two things with a property name. In English
>>  one makes an assertion by writing a sentence. Grammatical sentences
>>  must contain a main verb. It seems natural to link the verb with the
>>  property. It is impossible to create a 'natural' English rendering
>>  which has no verbs at all.
>In the RoleNoun pattern, the verb is a generic/implicit 'has';
>rather than
>   Dan knows Pat.
>we write
>   Dan has acquaintance Pat.
>In N3, we allow the 'has' to be left out:
>   Dan acquaintance Pat.

Well, I rest my case. If this is natural English then Im a Dutchman. 
And try this for properties corresponding to more active 
relationships, eg authorship. Start with Pat wrote/authored 
RDFSemantics, and keep the subject/object ordering. What word do you 
use after the has?

Hmm, actually you could say 'has written', I guess, but I think that 
is cheating. It inverts to 'RDFSemantics is written of Pat'  which 
has the wrong preposition, but I guess one could get used to it (?). 
Or you could just allow 'is <P> by'  as an alternative inversion 

>And we get the inverse for free using is/of keywords:
>   Pat is acquaintance of Dan.
>no need for a separate property for the inverse:
>   Pat isKnownBy Dan.
>>  >  The UN/CEFACT naming standards for instance, use nouns in
>>  >all cases.  My concern is that this community's singular use of 
>>verbs, whether
>>  >compounded with nouns or not, continues to unnecessarily alienate 
>>the rest of
>>  >the development world.
>>  RDF is intended for use by a vastly larger audience, however.
>>  >  To me, this is a topic that needs to be addressed by the
>>  >W3C because it was RDF's original specifications which started this truly
>>  >onerous practice without a shred of rationale.
>>  To me, the restriction to nouns would need some supporting rationale.
>>  How can one express a proposition using only nouns? My mother's name
>>  is Betty. What does one say? Betty Motherhood Pat?
>Pat has mother Betty.

What is the logic of why it should be this way round? I was writing 
the mother last, not first. Why did you invert it?

>which harks to javascript/C++ a la
>  Pat.mother = Betty.

True, but javascript hardly qualifies as a human natural language. 
And I don't really see most properties as selector functions.

>>  >  And the paper cited above hardly
>>  >finds that a standard practice exists for naming RDF properties
>>  There shouldn't be a standard practice for naming properties, IMO.
>>  There are just too many properties around.
>>  >--  gee, both
>>  >RDF and OWL themselves even have issues in this regard!
>>  What issues?
>subPropertyOf is ugly.

Oh well, yes, lots of it is UGLY. intersectionOf instead of 'and' is 
ugly. Nobody in their right mind would say 'subproperty' in English. 
The way to render foo subPropertyOf baz is to say that foos are (or: 
must be, or: are always) bazes, which is roughly what you would get 
by Engishifying the underlying logical form. But I don't see anything 
in there about noun/verb.

BTW, the way we render OWL in COE is pretty simple. We use labels on 
nodes and arcs of a graph, but you can read along them as a kind of 
pseudo-English, and then you get things like

Beaujolais is RedWine.  (subClassOf)

Beaujolais is defined as: RedWine, and whereGrown must be CentralRegion.

(Class defined by intersectionOf with ValueOf restriction. The colon 
indicates a branch point in the graph - an RDF list in OWL/RDF - and 
the comma is put at the end of each branch, ie after each list 
element.  So this encodes an RDF collection. )

Human is hasFingers must be exactly ten.  (subclass of numerical restriction)

The 'is' in the last example isn't very natural, of course, but it 
could be omitted in an English-style rendering, provided one is 
willing to put up with 'is defined as' or similar to indicate the 
(less common) necessary-and-sufficient case.

We did however give up trying when it came to subpropertyof, I have 
to admit; and we use graphic conventions rather than text to indicate 
domain and range and other technical stuff. I'd love to see how 
anyone proposes to state domains and ranges in English.


>>  Pat Hayes
>>  >
>>  >John McClure
>Dan Connolly, W3C http://www.w3.org/People/Connolly/
>D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E

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 19:59:02 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:58:14 UTC