Limitations of OWL 1.1 to RDF mapping

[Holger Knublauch-2]

The current OWL 1.1 Mapping to RDF Graphs draft [1] states

"Not every OWL 1.1 ontology can be serialized in RDF. In particular,
ontologies using the following features of OWL 1.1 cannot be serialized:

1. punning and
2. annotations on axioms."

Could anyone please clarify the implications of this statement.  Is the
plan of the group to leave it like this, or are changes to the OWL 1.1
spec underway to ensure that there will be a complete mapping to RDF
Graphs?  Or, will we see an "OWL 1.0.9" that will be complete in RDF but
with less features than 1.1?

Thanks,
Holger

[1] http://owl1_1.cs.manchester.ac.uk/rdf_mapping.html

[Bijan Parsia]

On Nov 15, 2006, at 9:52 PM, Holger Knublauch wrote:

> The current OWL 1.1 Mapping to RDF Graphs draft [1] states
>
> "Not every OWL 1.1 ontology can be serialized in RDF. In  
> particular, ontologies using the following features of OWL 1.1  
> cannot be serialized:

These statements should be read as qualified with "under the current  
mapping".

> 1. punning and
> 2. annotations on axioms."
>
> Could anyone please clarify the implications of this statement.  Is  
> the plan of the group to leave it like this, or are changes to the  
> OWL 1.1 spec underway to ensure that there will be a complete  
> mapping to RDF Graphs?  Or, will we see an "OWL 1.0.9" that will be  
> complete in RDF but with less features than 1.1?

The RDF mapping has lagged behind the others, but the plan is to  
extend the mapping to cover these cases.

Cheers,
Bijan.

[Dave Reynolds-3]

Bijan Parsia wrote:

>
> On Nov 15, 2006, at 9:52 PM, Holger Knublauch wrote:
>
>> The current OWL 1.1 Mapping to RDF Graphs draft [1] states
>>
>> "Not every OWL 1.1 ontology can be serialized in RDF. In particular,
>> ontologies using the following features of OWL 1.1 cannot be serialized:
>
> These statements should be read as qualified with "under the current
> mapping".
>
>> 1. punning and
>> 2. annotations on axioms."
>>
>> Could anyone please clarify the implications of this statement.  Is
>> the plan of the group to leave it like this, or are changes to the OWL
>> 1.1 spec underway to ensure that there will be a complete mapping to
>> RDF Graphs?  Or, will we see an "OWL 1.0.9" that will be complete in
>> RDF but with less features than 1.1?
>
> The RDF mapping has lagged behind the others, but the plan is to extend
> the mapping to cover these cases.

Presumably this will remain a purely syntactic mapping and the comment
in your "Next Steps for OWL" paper [1]:

    "A triple syntax is being provided for OWL 1.1, syntactically
    compatible with the triple syntax for OWL DL. However, for the above
    reasons, this syntax could not be given a meaning compatible with the
    RDF meaning for triples ..."

will still apply?

Dave

[1] http://owl-workshop.man.ac.uk/acceptedLong/submission_11.pdf

[Pat Hayes]

>Bijan Parsia wrote:
>>
>>On Nov 15, 2006, at 9:52 PM, Holger Knublauch wrote:
>>
>>>The current OWL 1.1 Mapping to RDF Graphs draft [1] states
>>>
>>>"Not every OWL 1.1 ontology can be serialized in RDF. In
>>>particular, ontologies using the following features of OWL 1.1
>>>cannot be serialized:
>>
>>These statements should be read as qualified with "under the
>>current mapping".
>>
>>>1. punning and
>>>2. annotations on axioms."
>>>
>>>Could anyone please clarify the implications of this statement.
>>>Is the plan of the group to leave it like this, or are changes to
>>>the OWL 1.1 spec underway to ensure that there will be a complete
>>>mapping to RDF Graphs?  Or, will we see an "OWL 1.0.9" that will
>>>be complete in RDF but with less features than 1.1?
>>
>>The RDF mapping has lagged behind the others, but the plan is to
>>extend the mapping to cover these cases.
>
>Presumably this will remain a purely syntactic mapping and the
>comment in your "Next Steps for OWL" paper [1]:
>
>    "A triple syntax is being provided for OWL 1.1, syntactically
>    compatible with the triple syntax for OWL DL. However, for the above
>    reasons, this syntax could not be given a meaning compatible with the
>    RDF meaning for triples ..."
>
>will still apply?

Actually that statement is false. Such a syntax COULD be given a
meaning compatible with the RDF semantics, and also (in a sense) with
the OWL 1.1 semantics; but it would be a stronger assertion than the
strict OWL 1.1 meaning. It is what you would get by merging the three
OWL 1.1 notions of identity into a single meaning (equivalent then to
owl:sameAs). Put another way, it is what you would get if OWL 1.1
used punning on its own notion of identity in the same way as it uses
it between classes/individuals/properties; which of course it does
not. If it did, many more entailments would be supported; but all OWL
1.1 entailments would be correct. Thus, an OWL 1.1 reasoner applied
to the RDF triple syntax understood as legal RDF would be a valid but
incomplete reasoner. Which is fine in my book: incomplete but valid
reasoners are often very useful. I suspect that anyone who uses OWL
1.1 with the intention of expressing genuine meta-modelling, instead
of the rather weak imitation provided by the punning semantics,

BTW, the statement later in the paper,

"The possibility of continuing along this line [that is, the RDF
style of semantics] in OWL 1.1 is called even more into question by
the impossibility of extending it to Semantic Web languages with
expressive power on a par with that of First-Order Logic"

is also false, or at best highly misleading. The ISO Common Logic
draft standard (http://cl.tamu.edu/#cl) uses the same basic semantic
construction as that used in RDF, i.e. unrestricted use of names,
without punning, in a full first-order framework, in exact harmony
with the semantics of RDF. Although ISO CL is unusually liberal in
its syntax, the actual logic (without sequence markers) is
first-order logic by all accepted semantic criteria (for example, it
satisfies compactness and Skolem-Lowenheim), and can be processed
using conventional first-order inference engines. This work has been
publicly available for open comment now for over three years, and has
been commented on by at least some of the authors of this paper. A
more recent extension called IKL
(http://www.ihmc.us/users/phayes/IKL/GUIDE/GUIDE.html ) continues
(and utilizes) this semantic construction to extend the expressivity
considerably beyond first-order logic, and is apparently known to at
least some of the authors, as they have asked me about it personally.
I can therefore only surmise that the reiteration of this falsehood
in a recent publication is a deliberate attempt to mislead readers,
in what appears to be a continuing and systematic attempt to diss
RDF. I am at a loss to understand the motivation for this apparently
self-destructive behavior.

Pat Hayes

>
>Dave
>
>[1] http://owl-workshop.man.ac.uk/acceptedLong/submission_11.pdf


--
---------------------------------------------------------------------
IHMC [phone] or [phone]   home
40 South Alcaniz St. [phone]   office
Pensacola [phone]   fax
FL 32502 [phone]    cell
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes

[Peter F. Patel-Schneider]

From: "Pat Hayes" <phayes@...>
Subject: Re: Limitations of OWL 1.1 to RDF mapping
Date: Fri, 17 Nov 2006 13:31:09 -0600

[...]

> BTW, the statement later in the paper,
>
> "The possibility of continuing along this line [that is, the RDF
> style of semantics] in OWL 1.1 is called even more into question by
> the impossibility of extending it to Semantic Web languages with
> expressive power on a par with that of First-Order Logic"
>
> is also false, or at best highly misleading. The ISO Common Logic
> draft standard (http://cl.tamu.edu/#cl) uses the same basic semantic
> construction as that used in RDF, i.e. unrestricted use of names,
> without punning, in a full first-order framework, in exact harmony
> with the semantics of RDF. Although ISO CL is unusually liberal in
> its syntax, the actual logic (without sequence markers) is
> first-order logic by all accepted semantic criteria (for example, it
> satisfies compactness and Skolem-Lowenheim), and can be processed
> using conventional first-order inference engines. This work has been
> publicly available for open comment now for over three years, and has
> been commented on by at least some of the authors of this paper. A
> more recent extension called IKL
> (http://www.ihmc.us/users/phayes/IKL/GUIDE/GUIDE.html ) continues
> (and utilizes) this semantic construction to extend the expressivity
> considerably beyond first-order logic, and is apparently known to at
> least some of the authors, as they have asked me about it personally.
> I can therefore only surmise that the reiteration of this falsehood
> in a recent publication is a deliberate attempt to mislead readers,
> in what appears to be a continuing and systematic attempt to diss
> RDF. I am at a loss to understand the motivation for this apparently
> self-destructive behavior.
>
> Pat Hayes

The above quote has the following context:

    A triple syntax is being provided for OWL 1.1, syntactically compatible
    with the triple syntax for OWL DL. However, for the above reasons, this
    syntax could not be given a meaning compatible with the RDF meaning for
    triples, at least not without some very difficult semantic tiptoeing
    (as well as some questionable encoding (such as creating fresh URI
    references for punning purposes, e.g., using Person-the-Class and
    Person-the-Individual instead of just Person). The appropriateness of
    continuing along this line with OWL 1.1 is called even more into
    question by the impossibility of extending it to Semantic Web languages
    with expressive power on a par with that of First-Order Logic [15].
   
    [Next Steps for OWL, Bernardo Cuenca Grau, Ian Horrocks, Bijan Parsia,
    Peter Patel-Schneider and Ulrike Sattler, 2006 OWL Experiences and
    Directions workshop, Athens, Georgia, 10-11 November 2006, p. 6,
    currently available at
    owl-workshop.man.ac.uk/acceptedLong/submission_11.pdf]

This context indicates that the impossibility involves maintaining key
aspects of RDF, including 1/ all syntax is triples and 2/ all triples have
at least their basic RDF meaning as facts.  As the ISO Common Logic draft
standard does not have these aspects it can hardly be used as a
counterexample to the claim.

Peter F. Patel-Schneider
Bell Labs Research

[Pat Hayes]

>From: "Pat Hayes" <phayes@...>
>Subject: Re: Limitations of OWL 1.1 to RDF mapping
>Date: Fri, 17 Nov 2006 13:31:09 -0600
>
>[...]
>
>>  BTW, the statement later in the paper,
>>
>>  "The possibility of continuing along this line [that is, the RDF
>>  style of semantics] in OWL 1.1 is called even more into question by
>>  the impossibility of extending it to Semantic Web languages with
>>  expressive power on a par with that of First-Order Logic"
>>
>>  is also false, or at best highly misleading. The ISO Common Logic
>>  draft standard (http://cl.tamu.edu/#cl) uses the same basic semantic
>>  construction as that used in RDF, i.e. unrestricted use of names,
>>  without punning, in a full first-order framework, in exact harmony
>>  with the semantics of RDF. Although ISO CL is unusually liberal in
>>  its syntax, the actual logic (without sequence markers) is
>>  first-order logic by all accepted semantic criteria (for example, it
>>  satisfies compactness and Skolem-Lowenheim), and can be processed
>>  using conventional first-order inference engines. This work has been
>>  publicly available for open comment now for over three years, and has
>>  been commented on by at least some of the authors of this paper. A
>>  more recent extension called IKL
>>  (http://www.ihmc.us/users/phayes/IKL/GUIDE/GUIDE.html ) continues
>>  (and utilizes) this semantic construction to extend the expressivity
>>  considerably beyond first-order logic, and is apparently known to at
>>  least some of the authors, as they have asked me about it personally.
>>  I can therefore only surmise that the reiteration of this falsehood
>>  in a recent publication is a deliberate attempt to mislead readers,
>>  in what appears to be a continuing and systematic attempt to diss
>>  RDF. I am at a loss to understand the motivation for this apparently
>>  self-destructive behavior.
>>
>>  Pat Hayes
>
>The above quote has the following context:
>
>     A triple syntax is being provided for OWL 1.1, syntactically compatible
>     with the triple syntax for OWL DL. However, for the above reasons, this
>     syntax could not be given a meaning compatible with the RDF meaning for
>     triples, at least not without some very difficult semantic tiptoeing
>     (as well as some questionable encoding (such as creating fresh URI
>     references for punning purposes, e.g., using Person-the-Class and
>     Person-the-Individual instead of just Person). The appropriateness of
>     continuing along this line with OWL 1.1 is called even more into
>     question by the impossibility of extending it to Semantic Web languages
>     with expressive power on a par with that of First-Order Logic [15].
>    
>     [Next Steps for OWL, Bernardo Cuenca Grau, Ian Horrocks, Bijan Parsia,
>     Peter Patel-Schneider and Ulrike Sattler, 2006 OWL Experiences and
>     Directions workshop, Athens, Georgia, 10-11 November 2006, p. 6,
>     currently available at
>     owl-workshop.man.ac.uk/acceptedLong/submission_11.pdf]
>
>This context indicates that the impossibility involves maintaining key
>aspects of RDF, including 1/ all syntax is triples and 2/ all triples have
>at least their basic RDF meaning as facts.  As the ISO Common Logic draft
>standard does not have these aspects it can hardly be used as a
>counterexample to the claim.

I was referring to the claim in the last sentence
of the above, which refers to first-order logic
and uses the word "impossibility", and refers to
the RDF style of SEMANTICS (my emphasis). That
sentence is rather hard to parse, however (what
exactly is the phrase "along this line" intended
to refer to? I assumed, after some thought, that
it refers to the "line" of NOT using punning,
rather than the "line" adopted in OWL 1.1) and I
may have misunderstood it: if so, I apologize.

BTW, I am not defending the idea that 'all syntax
is triples'; but that point is moot, since the
entire thread is predicated on the assumption
that at least the SYNTAX will indeed be triples.

Pat

>Peter F. Patel-Schneider
>Bell Labs Research


--
---------------------------------------------------------------------
IHMC [phone] or [phone]   home
40 South Alcaniz St. [phone]   office
Pensacola [phone]   fax
FL 32502 [phone]    cell
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes

[Ian Horrocks]

On 17 Nov 2006, at 17:01, Dave Reynolds wrote:

>
> Bijan Parsia wrote:
>> On Nov 15, 2006, at 9:52 PM, Holger Knublauch wrote:
>>> The current OWL 1.1 Mapping to RDF Graphs draft [1] states
>>>
>>> "Not every OWL 1.1 ontology can be serialized in RDF. In particular,
>>> ontologies using the following features of OWL 1.1 cannot be
>>> serialized:
>> These statements should be read as qualified with "under the current
>> mapping".
>>> 1. punning and
>>> 2. annotations on axioms."
>>>
>>> Could anyone please clarify the implications of this statement.  Is
>>> the plan of the group to leave it like this, or are changes to the
>>> OWL 1.1 spec underway to ensure that there will be a complete
>>> mapping to RDF Graphs?  Or, will we see an "OWL 1.0.9" that will be
>>> complete in RDF but with less features than 1.1?
>> The RDF mapping has lagged behind the others, but the plan is to
>> extend the mapping to cover these cases.
>
> Presumably this will remain a purely syntactic mapping and the comment
> in your "Next Steps for OWL" paper [1]:
>
>    "A triple syntax is being provided for OWL 1.1, syntactically
>    compatible with the triple syntax for OWL DL. However, for the above
>    reasons, this syntax could not be given a meaning compatible with
> the
>    RDF meaning for triples ..."
>
> will still apply?

The current status is that OWL 1.1 has a triple syntax that is fully
backwards compatible with OWL: any OWL DL ontology is an OWL 1.1
ontology. There are a couple of new features of OWL 1.1 that we were
having trouble figuring out how to serialise as triples, but after
discussions at ISWC and the OWLED workshop (thanks to Alan Ruttenberg
for some helpful suggestions) we now have a solution, and a revision of
the triple syntax that covers all of OWL 1.1 will soon be available.
Regarding the semantics, it may not be possible to extend OWL's
RDF-Compatible Model-Theoretic Semantics so as to *completely* capture
the meaning of OWL 1.1 while at the same time giving *all* the triples
their basic RDF meaning as facts. Due to the backwards compatibility of
the syntax, however, the meaning of the OWL part of OWL 1.1 will be
captured by the existing RDF-Compatible Semantics, and interest has
been expressed in trying to extend this (the existing RDF-Compatible
Semantics) so as to capture at least some of the meaning of the OWL 1.1
extensions.

Regards,

Ian



>
> Dave
>
> [1] http://owl-workshop.man.ac.uk/acceptedLong/submission_11.pdf
>
>

[Jim Hendler]

Ian - if I try to translate the below into a more informal form, I
think it says that OWL 1.1 will be an extension of OWL DL, but
possibly not of OWL Full.  Is that right?  WOuld there still be an
effort to make sure Full is enhanced with the same features that DL
is (the way the current DL and Full remain syntactically aligned)?
  thanks
  JH


At 3:35 PM +0000 11/20/06, Ian Horrocks wrote:

>On 17 Nov 2006, at 17:01, Dave Reynolds wrote:
>
>>
>>  Bijan Parsia wrote:
>>>  On Nov 15, 2006, at 9:52 PM, Holger Knublauch wrote:
>>>>  The current OWL 1.1 Mapping to RDF Graphs draft [1] states
>>>>
>>>>  "Not every OWL 1.1 ontology can be serialized in RDF. In
>>>>particular, ontologies using the following features of OWL 1.1
>>>>cannot be serialized:
>>>  These statements should be read as qualified with "under the
>>>current mapping".
>>>>  1. punning and
>>>>  2. annotations on axioms."
>>>>
>>>>  Could anyone please clarify the implications of this statement.
>>>>Is the plan of the group to leave it like this, or are changes to
>>>>the OWL 1.1 spec underway to ensure that there will be a complete
>>>>mapping to RDF Graphs?  Or, will we see an "OWL 1.0.9" that will
>>>>be complete in RDF but with less features than 1.1?
>>>  The RDF mapping has lagged behind the others, but the plan is to
>>>extend the mapping to cover these cases.
>>
>>  Presumably this will remain a purely syntactic mapping and the
>>comment in your "Next Steps for OWL" paper [1]:
>>
>>     "A triple syntax is being provided for OWL 1.1, syntactically
>>     compatible with the triple syntax for OWL DL. However, for the above
>>     reasons, this syntax could not be given a meaning compatible with the
>>     RDF meaning for triples ..."
>>
>>  will still apply?
>
>The current status is that OWL 1.1 has a triple syntax that is fully
>backwards compatible with OWL: any OWL DL ontology is an OWL 1.1
>ontology. There are a couple of new features of OWL 1.1 that we were
>having trouble figuring out how to serialise as triples, but after
>discussions at ISWC and the OWLED workshop (thanks to Alan
>Ruttenberg for some helpful suggestions) we now have a solution, and
>a revision of the triple syntax that covers all of OWL 1.1 will soon
>be available. Regarding the semantics, it may not be possible to
>extend OWL's RDF-Compatible Model-Theoretic Semantics so as to
>*completely* capture the meaning of OWL 1.1 while at the same time
>giving *all* the triples their basic RDF meaning as facts. Due to
>the backwards compatibility of the syntax, however, the meaning of
>the OWL part of OWL 1.1 will be captured by the existing
>RDF-Compatible Semantics, and interest has been expressed in trying
>to extend this (the existing RDF-Compatible Semantics) so as to
>capture at least some of the meaning of the OWL 1.1 extensions.
>
>Regards,
>
>Ian
>
>
>
>>
>>  Dave
>>
>>  [1] http://owl-workshop.man.ac.uk/acceptedLong/submission_11.pdf
>>
>>

--
Prof James Hendler hendler@...
Dept of Computer Science http://www.cs.umd.edu/~hendler
AV Williams Bldg 301-405-2696 (work)
Univ of Maryland 301-405-6707 (Fax)
College Park, MD 20853 USA

[Ian Horrocks]

On 20 Nov 2006, at 17:28, Jim Hendler wrote:

>
> Ian - if I try to translate the below into a more informal form, I
> think it says that OWL 1.1 will be an extension of OWL DL, but
> possibly not of OWL Full.  Is that right?


Up to now, there hasn't been any attempt to extend the RDF-Compatible
Model-Theoretic Semantics so as to capture the meaning of OWL 1.1.
Assuming that this is what you mean by not being an extension of OWL
Full, then it is certainly right that this is the current status.

> WOuld there still be an effort to make sure Full is enhanced with the
> same features that DL is (the way the current DL and Full remain
> syntactically aligned)?

Of course every OWL 1.1 ontology is an OWL Full ontology, so in this
sense OWL Full, by its very nature, will be syntactically aligned with
OWL 1.1. I guess what you are asking about is extending the
RDF-Compatible Model-Theoretic Semantics so as to make the extra syntax
meaningful in OWL Full. As I mentioned, interest has been expressed in
working on this, so I guess that the answer is yes -- it is likely that
there will be such an effort.

Ian

>  thanks
>  JH
>
>
> At 3:35 PM +0000 11/20/06, Ian Horrocks wrote:
>> On 17 Nov 2006, at 17:01, Dave Reynolds wrote:
>>
>>>
>>>  Bijan Parsia wrote:
>>>>  On Nov 15, 2006, at 9:52 PM, Holger Knublauch wrote:
>>>>>  The current OWL 1.1 Mapping to RDF Graphs draft [1] states
>>>>>
>>>>>  "Not every OWL 1.1 ontology can be serialized in RDF. In
>>>>> particular, ontologies using the following features of OWL 1.1
>>>>> cannot be serialized:
>>>>  These statements should be read as qualified with "under the
>>>> current mapping".
>>>>>  1. punning and
>>>>>  2. annotations on axioms."
>>>>>
>>>>>  Could anyone please clarify the implications of this statement.
>>>>> Is the plan of the group to leave it like this, or are changes to
>>>>> the OWL 1.1 spec underway to ensure that there will be a complete
>>>>> mapping to RDF Graphs?  Or, will we see an "OWL 1.0.9" that will
>>>>> be complete in RDF but with less features than 1.1?
>>>>  The RDF mapping has lagged behind the others, but the plan is to
>>>> extend the mapping to cover these cases.
>>>
>>>  Presumably this will remain a purely syntactic mapping and the
>>> comment in your "Next Steps for OWL" paper [1]:
>>>
>>>     "A triple syntax is being provided for OWL 1.1, syntactically
>>>     compatible with the triple syntax for OWL DL. However, for the
>>> above
>>>     reasons, this syntax could not be given a meaning compatible
>>> with the
>>>     RDF meaning for triples ..."
>>>
>>>  will still apply?
>>
>> The current status is that OWL 1.1 has a triple syntax that is fully
>> backwards compatible with OWL: any OWL DL ontology is an OWL 1.1
>> ontology. There are a couple of new features of OWL 1.1 that we were
>> having trouble figuring out how to serialise as triples, but after
>> discussions at ISWC and the OWLED workshop (thanks to Alan Ruttenberg
>> for some helpful suggestions) we now have a solution, and a revision
>> of the triple syntax that covers all of OWL 1.1 will soon be
>> available. Regarding the semantics, it may not be possible to extend
>> OWL's RDF-Compatible Model-Theoretic Semantics so as to *completely*
>> capture the meaning of OWL 1.1 while at the same time giving *all*
>> the triples their basic RDF meaning as facts. Due to the backwards
>> compatibility of the syntax, however, the meaning of the OWL part of
>> OWL 1.1 will be captured by the existing RDF-Compatible Semantics,
>> and interest has been expressed in trying to extend this (the
>> existing RDF-Compatible Semantics) so as to capture at least some of
>> the meaning of the OWL 1.1 extensions.
>>
>> Regards,
>>
>> Ian
>>
>>
>>
>>>
>>>  Dave
>>>
>>>  [1] http://owl-workshop.man.ac.uk/acceptedLong/submission_11.pdf
>>>
>>>
>
> --
> Prof James Hendler hendler@...
> Dept of Computer Science http://www.cs.umd.edu/~hendler
> AV Williams Bldg 301-405-2696 (work)
> Univ of Maryland 301-405-6707 (Fax)
> College Park, MD 20853 USA