DTTF: Embracing the dark side

In the shadow of Peter's message [1] I am beginning to go over to
the dark side.

Here is how I see the argument.

====

DAML+OIL Model Theory shows how an RDF graph can be read as a
concrete syntax for a description logic.
Description logics distinguish between the Terminology in the T-BOX
which describe the ontology, and the A-BOX which describes the
instance data.


Such a reading allows entailments like:

A:
[[
Premise 
 John rdf:type Student .
 John rdf:type Employee .

Conclusion 
 John rdf:type _:i .
 _:i rdf:type daml:Class .
 _:i daml:intersectionOf _:l .
 _:l rdf:type daml:List .
 _:l daml:first Student .
 _:l daml:rest _:t .
 _:t rdf:type daml:List .
 _:t daml:first Employee .
 _:t daml:rest daml:nil .
]]

WOWG wishes that OWL support such entailments.

Viewing DAML as a surface syntax for description logic
involves regarding some of the triples as part of the 
T-BOX and some of the triples as part of the A-BOX.
Wihin DL, and hence within this view of DAML+OIL, 
T-BOX reasoning and A-BOX reasoning are distinct
so, for example when

B:
[[

 foo rdfs:subPropertyOf bar .

is in the T-BOX

and

 a  foo b .

is in the A-BOX

then we can infer that

  a bar b .

is in the A-BOX.
]]

In contrast, Within RDFS there is no such distinction
so for example:

C:
[[
  foo rdfs:subPropertyOf daml:oneOf .
  a foo b .

rdfs-entails

  a daml:oneOf b .
]]

Entailments like this cannot be handled by the DAML
model theory. (Because the entailed triple is in the
T-BOX, but the T-BOX is understood as a *syntactic*
representation of the DL abstract syntax).

WOWG has considered approaches that work with (C) 
[4], [5]; but they do not explain entailment (A).
A further approach [6] that tried to resolve this
proved to require more research [1].
  
Hence OWL is unlikely to respect entailment (C).

One approach is for OWL to treat the RDF graph as a
syntax for the T-BOX, and to respect the RDF(S) semantics
for the A-BOX. When the RDF(S) semantics impinges on the 
OWL T-BOX such arcs are not respected by the OWL 
semantics.
A slightly different variant is simply to say that
RDFS that would modify the OWL T-BOX is not legal
OWL. In this variant the premise of (C) is not an OWL
document.

Either way, there is a divergence between the OWL
semantics and the RDFS semantics.

Another aspect of the strict separation between
A-BOX and T-BOX and the reuse of DL semantics
is that the class expressions and relationships
between them, are *not* part of the domain of 
discourse, unlike in RDFS. Thus the triples involved
in such class expressions do not necessarily conform
with RDF semantics.

In summary, a known to work approach to the
semantics of DAML+OIL, that does embrace
entailment (A), is to reuse DL semantics.
This approach requires dark triples to be used
for the T-BOX information (i.e. most/all of
the owl namespace, and maybe some of the rdfs
namespace).




===

Personally I find that this argument motivates me to 
attempt the research required under [1]; but such work 
is not appropriate for the WG.



Jeremy 




[1] Peter's problem
http://lists.w3.org/Archives/Public/www-webont-wg/2002May/0064.html


[4] axiomatic semantics of DAML+OIL

[5] solipisitic
http://lists.w3.org/Archives/Public/www-webont-wg/2002Mar/0179.html
[6] my attempt
http://lists.w3.org/Archives/Public/www-webont-wg/2002May/0061.html
http://lists.w3.org/Archives/Public/www-webont-wg/2002Apr/0155.html

Received on Tuesday, 14 May 2002 08:38:45 UTC