RE: DTTF: How unasserted triples help

>-----Original Message-----
>From: pat hayes [mailto:phayes@ai.uwf.edu]
>Sent: Thursday, 23 May, 2002 21:47
>To: Dan Connolly
>Cc: Jonathan Borden; WebOnt WG
>Subject: Re: DTTF: How unasserted triples help
>
>
>>>  First using "asserted" triples, the following:
>>>
>>>  intersectionOf( student, employee ) or in N3
>>>
>>>  _:c1 rdf:type owl:Class
>>>  _:c1 owl:intersectionOf _:L1
>>>  _:L1 rdf:type owl:List
>>>  _:L1 owl:first <#student>
>>>  _:L1 owl:rest _:L2
>>>  _:L2 rdf:type owl:List
>>>  _:L2 owl:first <#employee>
>>>  _:L2 owl:rest owl:nil
>>>
>>>  which roughly translates to: "There exists a class having 
>the intersectionOf
>>>  property whose object is a List, whose first element is 
>#student and whose
>>>  rest is another List whose first element is #employee and 
>whose rest is nil"
>>>
>>>  As "asserted" triples, these statements are said to be 
>_truths_, which is a
>>>  bit strange because this isn't what we really mean.
>>
>>It's not strange at all; it's exactly what we mean:
>>
>>"for intersectionOf(X, Y) read: X is the intersection of the 
>classes in
>>the list Y"
>>  -- http://www.w3.org/2001/10/daml+oil
>
>Well then, please someone write a model theory for DAML which makes 
>it have this meaning. The DAML MT does not refer to lists.  English 
>paraphrases are not reliable indicators of precise meaning.
>
>For the record, I have never understood DAML to mean this (ie that 
>daml:intersectionOf asserts that a list exists in the domain of 
>discourse) and find this a slightly crazy way to interpret the 
>language. In fact, if this really is the intended meaning of 
>DAML+OIL, then I would urge that the decision to base OWL on DAML+OIL 
>should be reconsidered at the earliest opportunity, before we waste 
>any more time and effort.
>
>>  > Instead we mean to say:
>>>
>>>  "There exists a class which is the intersectionOf #student 
>and #employee."
>>
>>That's an informal corrollary of the above, but it can't be
>>stated formally/directly using 2-place predicates.
>
>True; binary relations are very limited. It can be expressed using 
>binary functions, however. In KIF: (equal A (intersect B C)). Both 
>intersect and equality are binary.



But this is (part) of the point, isn't it? That one is lead to asserting
the existence of lists in the domain of discourse because of the
syntactic bounds of the language. The reasoning goes as follows:

1) Anything and everything one wishes to express in RDF syntax must be
done by way of 2-place predicates.
2) So expressing the most elementary facts about intersections leads us
to these strange uses of lists, and assertions involving lists
3) Use that, along with desired entailments and RDF constructions to
build the paradoxes with which we have become so familiar in this
working group


Avoiding the paradoxes is attained basically by way of avoiding the
assertion of these troublesome lists. Some ways of attaining this:

a) Permitting expressions that go beyond binary relations (such as the
binary functions above) to avoid the use of lists
b) Thinking up some creative manner to express intersection with binary
relations but without using lists 
c) Using dark (unasserted) triples for expressing lists

 
Lacking a creative version of b), and rejecting a) as going against the
spirit of RDF, many have suggested adopting c). But there are many
others who are unhappy about c), as well. 



-- Ziv


PS Yes I have returned to daily life -- and to participating in this
working group -- safe and sound.

Received on Monday, 27 May 2002 12:06:46 UTC