Minutes of Sept. 5 Telecon

Minutes of Sept. 5, 2002 Telecon

WEB ONTOLOGY WORKING GROUP
September 5, 2002

Quotes and attributions are approximate.
See chat log for details http://www.w3.org/2002/09/5-webont-irc


Chair: Guus Schreiber
Scribe: Leo Obrst

SUMMARY:
1) RESOLUTIONS

RESOLVED: FTF5 Will be held in Manchester


2) NEW ACTIONS:

From 1.2:
ACTION: Revision of Aug. 29 minutes to  be done by Tim Finin by next
week and to be resolved next week.
See: Description of minutes: 27 Aug. Tips for Scribes by Dan Connolly.

From 1.4:
ACTION: Ian Horrocks will look at alternate arrangements for vtc (in
Manchester for F2F5).

From 3.1, Issue 4.4:
ACTION: Mike Dean: will document how he's using RDF statement tags, and
how to solve this. 
Does his solution satisfy the requirements in the requirements document?

From 3.1, Issues 5.6, 5.14:
ACTION: Proposal by J. Heflin to couple 5.6 and 5.14, and will be
separate section in reference document

From 3.2, Issue 5.1:
ACTION: Frank will come up with some more use cases (ontology merging).
ACTION: Ian to make a suggestion on how to solve this, adding properties
to data values
ACTION: Dan to elaborate his use cases/requirements on this.  

From 3.2, Issue 5.19:
ACTION: Pat and Jeremy to create this as an issue and create a sketch of
an idea, by next Tuesday.
ACTION: Peter: the status quo meets the requirement: give the same name
to a class and an instance. Show the mechanism or point to a use case to
the old discussion.

3.3 Issue 5.3:
ACTION: by COB Monday, Pat to complete first document  (and Peter and
Pat to discuss this on phone to next telecon). Ian will send comments to
Peter and Pat. 
ACTION: Pat will send out (Great Horned Owl) document by next telecon.

4. Guide:
ACTION: Guus, MOVE THIS ISSUE TO AGENDA ITEM FOR NEXT WEEK.



DETAIL:

1) ADMIN (15 min)

1.1 Roll call

Present:

Jonathan Borden
Jeremy Carroll
Dan Connolly 
Jos De Roo
Mike Dean 
Larry Eshelman
Tim Finin 
Nicholas Gibbins 
Pat Hayes 
Jeff Heflin 
Ziv Hellman 
James Hendler
Ian Horrocks
Ruediger Klein 
Ora Lassila 
Deborah McGuinness
Libby Miller 
Leo Obrst 
Peter Patel-Schneider 
Guus Schreiber  
Michael Smith 
John Stanton 
Lynn Andrea Stein 
Herman ter Horst 
Lynne R. Thompson 
Frank van Harmelen 
Evan Wallace 

Regrets: Dale, Motta, Sabbouh,
Volz, Marchiori, Welty, Yanosy

1.2 Minutes previous meeting

PROPOSED to accept as minutes of the Aug 29 telecon:
http://lists.w3.org/Archives/Public/www-webont-wg/2002Sep/0047.html

Peter Patel-Schneider objects. Does not include some action items.
Does not record the walkthrough action.
Current agenda: Does not record that Mike Dean will be doing the
comparison document.

Vote regarding OWL test of Aug. 29: 
Opposed by Peter Patel-Schneider
Abstention: Chris Welty

ACTION: Revision of Aug. 29 minutes to  be done by Tim Finin by next
week and to be resolved next week.
See: Description of minutes: 27 Aug. Tips for Scribes by Dan Connolly.


1.3 Next telecon

- Sep 12, per regular schedule
- Scibe for Sep 12 is solicited: Jeremy 

1.4 Upcoming ftf meetings

Bristol ftf update (Jeremy):
http://www-uk.hpl.hp.com/people/jjc/webont/webontOct.html
ACTION: Jeremy will provide entertainment info on F2F4.
CONTINUED

Resolution on location of ftf5 (will be held on Jan 9-10 only if needed) 
based on results of vote (Dan):
http://www.w3.org/2000/11/msm/showv.php3?wgid=30310&qno=ftf5v

RESOLVED. Will be in MANCHESTER.
ACTION: Ian Horrocks will look at alternate arrangements for vtc (in
Manchester for F2F5s).


2) ACTION review: (5 min)


ACTION Larry Eshelman to contribute example to how-to-do-it doc
CONTINUED, will be done next week.

ACTION: Hendler produce test for issue 4.2
CONTINUED.

ACTION: Guus to update UML document to reflect new Owl Lite
CONTINUED, will be sent out next week. 

ACTION: Chairs to find an editor or process to produce the RDF/RDFS/OWL
Lite/OWL comparison/discussion document. JimH to inform SW-CG this is
being considered.
CONTINUED.

ACTION: DanC will hold email vote on location of ftf5: Manchester or
NY. One vote per org. seven days to respond
DONE.

ACTION: Mike Smith to distribute draft walkthru by Sep 4
DONE

ACTION: Mike Dean will be doing the comparison document.
CONTINUED

3) ISSUES (25-30 min)

3.1 Issues w.r.t additional language features

- short discussion on how to proceed
- if possible, assignment of issue owners

ISSUE 4.4 Extra-logical feature set
http://www.w3.org/2001/sw/WebOnt/webont-issues.html#I4.4-Extra-logical-feature-set

Dan: Jim H. is strong proponent of this.  
Frank: someone asked for human readable labels for classes. 
General purpose tagging mechanism subsumes this: who edited it, etc.

Dan: make tags on assertions? 
Allow what RDF allows: just tagging on objects?
How about tagging of statements? Mike Dean would do away with.
RDF statement tags ok for his purpose. 
Frank: the req. document has something more than RDF reification. 

ACTION: Mike Dean: will document how he's using RDF statement tags, and
how to solve this. 
Does his solution satisfy the requirements in the requirements document?

ISSUE 5.6 daml:imorts as magical syntax
http://www.w3.org/2001/sw/WebOnt/webont-issues.html#I5.6-daml:imports-as-magic-syntax

ISSUE 5.14 ontology versioning
http://www.w3.org/2001/sw/WebOnt/webont-issues.html#I5.14-Ontology-versioning

ACTION: Proposal by J. Heflin to couple 5.6 and 5.14, and will be
separate section in reference document


3.2 ISSUES 5.1 & 5.19

- short discussion on possible actions to move forward
See also message by chairs:
http://lists.w3.org/Archives/Public/www-webont-wg/2002Sep/0008.html

Jim and Guus looked at ISSUES list and those with impact on
tractability: 5.1, 5.19
Guus: any proposal for moving forward on these issues.
 
ISSUE 5.1: Uniform treatment of literal/data values
    [The DAML+OIL distinction between object- and datatypeproperties
    prevents use of e.g.   FunctionalProperty for datatype-properties]
http://www.w3.org/2001/sw/WebOnt/webont-issues.html#I5.1-Uniform-treatment-of-literal-data-values

Guus: functional properties like state code (Dan C's example). A valid
point, any possible actions for middle ground. 
Ian: reasonable use cases. The solution of removing this distinction
makes the language much too strong. Don't need this. Work of Ian with
Dresden folks: some properties act as a key for an object. This is what
the use cases really address.
Dan: class of odd numbers, etc.
Ian: can have these as datatypes
Dan: class of rational numbers?
Mike S: can apply datatypes to classes. Nothing to prevent this.
Ian: Dan wants to have class of 1 is a member of the class of ...
Mike S: but can say instance of 1 is ...
Pat: classes of numbers should be ok?
Subjects can't be literals.

Ian: key type property, but ongoing research issue
Frank: merging ontologies with disparate types/values
Pat: isn't this a problem, even if datatype-class is one to one?
Ian: can still have separate reasoners between datatypes and objects
(via oracle) and then pass inferences
Dan: Frank, any more use cases? 

ACTION: Frank will come up with some more use cases (ontology merging).

Hay: discomfort about implementation/efficiency issues affecting the
language
Ian: but then why didn't we go with a much stronger language to begin
with, in which we can do anything?

Ian: we started with DAML+OIL, we could start again from scratch (this
is a bug in DAML+OIL)
Dan: we recognized this as a bug.
Mike S: functional use of datatype properties seem to be solvable
without a uniform treatment

ACTION: Ian to make a suggestion on how to solve this, adding properties
to data values
ACTION: Dan to elaborate his use cases/requirements on this.  

    ISSUE 5.19: Classes as instances
    [Requirement 14; several use cases, RDF feature, but not clear how
OWL
    can support this. Also termed: "classes/properties should be in the
    domain of discourse" ]
http://www.w3.org/2001/sw/WebOnt/webont-issues.html#I5.19-Classes-as-instances

Pat H: OWL and RDF represent the extremes on this position. How about an
intermediate proposal.
Can have layers of universes, no loops (ramified classes?) What's the
appropriate set theory?
Is it known that being liberal with this question would break DL?
Ian: how to deal with this? Is this even in FOL?
Have lists to be in the domain of discourse. Lists of lists: not
countably infinite? Then not in FOL.
Pat H: no, countable. Pat has a paper in the RDF list on this. 

Guus: can you constrain the way you use classes?
Ian: may be possible, but by choosing this path, trying to solve some
hard research problem.
Pat H: sympathizes with this.  
Ian: what do you want to do with the language once you revise it.
Dan: I want to use a FOL theorem-prover on this
Peter:  Stickler's engine would break on this.
Dan: sequence quantifiers not needed
Peter: Stickler's dependent on individual and properties/predicates.
Ian: no FOL theorem-provers can handle this?
Pat: May be trivial change.
Ian: strange, restricted language that requires a FOL theorem-prover but
have no access to FOL?
Dan: all will have to make some syntactic changes, so this is trivial.
Ian: not trivial, more axiomatizations needed. Axiomatizations up to
this point are buggy.
Dan: DAML+OIL, ok.
Pat: to focus on the more tractable discussion, see his paper.
Guus: how do we move forward on this?
Ian: if we could offer choice: full FOL but not classes as instances, or
could but in a restricted language? 
Which satisfies the larger number of people? Probably the first. What
are the compelling use cases?
Allow instances and classes with same name. 
Jonathan: can define a subset of a reasoner that considered classes as
instances and one that didn't.

Jim Hendler just called in from cell.

Guus: need to move forward on this. 
Ian: sizes of the domain of discourse is a problem.
Pat: can't really offer syntactic solution of Jonathan. 
Pat: if the layers are syntactically marked.

ACTION: Pat and Jeremy to create this as an issue and create a sketch of
an idea, by next Tuesday.
ACTION: Peter: the status quo meets the requirement: give the same name
to a class and an instance. Show the mechanism or point to a use case to
the old discussion.


3.3 ISSUE 5.3 Semantic layering

- discussion on Pat's revised MT doc
http://lists.w3.org/Archives/Public/www-webont-wg/2002Aug/0236.html

Pat: major issue in earlier draft: nothing ensured that universe
contained right things, so he made sure that universe was closed in the
right way (Aug 27).  Another version: closure is functions (not yet
available).
One can ensure that universe contains the right things. 
The newer version is better and easier to follow.
Dan: can deal with Russell paradoxes?
Pat: yes, but doesn't handle "policy guards".


- short discussion on the required priority of the "dark-triple"
request to RDF Core


This request cam from discusion on issue 5.3:
http://www.w3.org/2001/sw/WebOnt/webont-issues.html#I5.3-Semantic-Layering

Peter: can't determine this yet, since we haven't seen Pat's document.
The one that Peter and Ian have seen: more than 10% objections. 

Will basic approach work? I.e., embedding OWL in RDF/S will work, but
not layering?

Guus: any other comments. Does this document seem a reasonable proposal?
Pat: must distinquish clearly between RDF: Class and OWL: Class
Peter: a small price to pay to getting this down.
Dan: OWL classes and instances are different. Can't talk about these in
RDF/S.

ACTION: by COB Monday, Pat to complete first document  (and Peter and
Pat to discuss this on phone to next telecon). Ian will send comments to
Peter and Pat. 
ACTION: Pat will send out (Great Horned Owl) document by next telecon.

DARK TRIPLES:
Pat: this is independent of semantic layering issue.
Jim: prioritization to RDF Core?
Jonathan: if leave at status quo, nothing is happening.
Jeremy: not actually vital. 
Jonathan: don't allow leakage, need formal group-to-group approach.
Jeremy: most promising is a solution that doesn' have dark triples, that
RDF Core can finish.

Peter: make clear that our request is still standing.


4 ) GUIDE (10-15 min: Mike S & others)

ACTION: Guus, MOVE THIS ISSUE TO AGENDA ITEM FOR NEXT WEEK.

- short discussion on first version walkthru (link to follow)
- actions for document completion
- release plan for other GUIDE docs (which ones should be ready by LC?)

5) Comments on WDs (10-15 min, editors)

Summary and short discussion on comments receibed sofar on the three
WDs.


6) AOB (0-5 min)

Received on Tuesday, 10 September 2002 09:47:25 UTC