Warning:
This wiki has been archived and is now read-only.

Teleconference.2008.01.30/Minutes

From OWL
Jump to: navigation, search

These minutes have been approved by the Working Group and are now protected from editing. (See IRC log of approval discussion.)

See also: IRC log


Rinke Hoekstra: ScribeNick bcuencag

(Scribe changed to Bernardo Cuenca Grau)

Admin

Ian Horrocks: Agenda amendments?

Ian Horrocks: No amendments, accept previous minutes?

Rinke Hoekstra: +1 to accept, they look fine to me
Martin Dzbor: +1 very comprehensive

Ian Horrocks: minutes approved

RESOLVED: approve previous minutes http://www.w3.org/2007/OWL/wiki/Teleconference.2008.01.23/Minutes

Doug Lenat: +1

Pending Actions

Ian Horrocks: pending actions

Ian Horrocks: Action 54

Ian Horrocks: Action 54 complete

Ian Horrocks: Action 57

Ian Horrocks: people should speak up if they think an action is not completed

Carsten Lutz: the action requested the definition of OWLPrime

Carsten Lutz: not clear what OWL Prime we should discuss

The document is the starting point for discussing OWL Prime

Carsten Lutz: for defining OWL Prime we still need some work

Ian Horrocks (guest): Action 67


Ian Horrocks: Action 71, Boris completed

Ian Horrocks: Action 74, dealt with by Jeremy

Ian Horrocks: Action 90: no agreement for acceptance

Ian Horrocks: Action 90 is ongoing

Bijan Parsia: Still outstanding
Bijan Parsia: Yes

Ian Horrocks: rich annotations action, any progress?

Bijan Parsia: yes

Ian Horrocks: Action 62

Alan Ruttenberg: action not yet completed

Ian Horrocks: use case for punning; postponed for next week

Ian Horrocks: Action 72, to be continued

Ian Horrocks: UML association; completed

Ian Horrocks: No proposals to resolve issues; more time for discussion


Discussion on Fragments

Ian Horrocks: fragments and conformance issues

Uli Sattler: yes - but couldn't find anything about "conformance"

Alan Ruttenberg: at the workshop we achieved a consensus


Alan Ruttenberg: we call fragments to syntax fragments and semantic fragments we called them conformance levels


Ian Horrocks: we agree that the email I sent is a reasonable starting point

Ian Horrocks: Do people understand the difference between fragments and conformance levels?

Carsten Lutz: happy with the distinction, but not clear what a conformance level is


Alan Ruttenberg: we could have a reasoner that does incomplete reasoning, but complete up to a certain set of entailments

Bijan Parsia: I suspect it's like what's talked about in: http://www.cs.mu.oz.au/research/mercury/information/doc-release/mercury_ref/Semantics.html#Semantics

James Hendler: we have a very tight definition of language fragments

Carsten Lutz: So they are orthogonal? We can have syntactic fragment X and conformance level Y at the same time?
Bijan Parsia: """However, implementations are also allowed to support other operational semantics, which may have non-determinism, so long as they are sound with respect to the declarative semantics, and so long as they meet a minimum level of completeness (they must be at least as complete as the strict commutative semantics, in the sense that every program which terminates for all possible orderings must also terminate in any implementation-defined operational semantics)."""

James Hendler: a fragment could be rather seen as a set of figures that could be supported

Uli Sattler: having conformance levels might be a good idea

Uli Sattler: it gives an idea of what it means to cover a certain construct

Uli Sattler: not too difficult to come up with a definition of conformance level

Carsten Lutz: I am not sure I got that

Uli Sattler: what it means to be correct or complete for a certain class of queries

Jeremy Carroll: +1 for test cases

Zhe Wu: are we going to provide a set of test cases and ensure that implementations should cover them?

Zhe Wu: does this relate to conformance?

Note (guest): the w3c generally doesn't do confromance certification [Scribe assist by Bijan Parsia]

Ian Horrocks: My assumption was that we would not define conformance in terms of test cases, but rather something more precise

Bijan Parsia: +1 to steering clear of WG doing certification

Sandro Hawke: OWL does have test cases

James Hendler: +1 to bijan's +1
Jeremy Carroll: FYGI see http://www.w3.org/TR/owl-test/#runningConsistencyChecker

Sandro Hawke: most standards in industry talk about what a language does

Sandro Hawke: OWL does not specify what a classifier should do

James Hendler: +1 to line up with defining software

Sandro Hawke: we should come up with a way of defining what a software does

Sandro Hawke: I push for conformance levels

Carsten Lutz: conformance level can be related to PD* semantics

Michael Smith: I wouldn't be confortable specifying fragments without taking into account the proper semantics

Uli Sattler: 1) it answers "Y is consistent" if and only if it is indeed consistent
Uli Sattler: 2) if it finds that "C1 is a subclass of C2", then it is indeed one (but not necessarily the other way round, and
Uli Sattler: 3) if we ask the reasoner the return all instances of a class C, then only such nstances are returned (but some might be missed)
Bijan Parsia: -1 to what alan said

Alan Ruttenberg: W3C does validation, and that could be done using tests

James Hendler: take sameAs

James Hendler: has a precise semantics and an RDF match

Boris Motik: To slightly refine Uli's idea, we probably need to look at both the syntactic level, and at the semantic level.
Boris Motik: You can have conformance first at the syntactic level.

James Hendler: we should tell the implementors what features they should implement

Boris Motik: Once you say what kind of syntax you accept, the semantic conformance level (a la Uli) would tell you what you are supposed to derive.

James Hendler: I'd like a fragment to be defined as a set of language features

James Hendler: one could implement OWL DL features but not following the DL semantics

Michael Smith: concerned uli's example is incomplete in IRC log, will retype first line
Michael Smith: here it is: E.g., we could say that a reasoner that is X-conformant behaves as follows when it handles an OWLPrime ontology Y:


Bijan Parsia: I do not know what validation is

Alan Ruttenberg: http://www.w3.org/RDF/Validator/
Alan Ruttenberg: http://www.w3.org/RDF/Validator/check?uri=referer

Bijan Parsia: defining a set of test cases would not be a suitable kind of validation

Zhe Wu: I have discussed about EL++ and Dl-Lite and DLP

Zhe Wu: DLP fits better with rules than EL++ or DL Lite

Zhe Wu: I like DLP better

Zhe Wu: Either PD* semantics or DLP would work in principle

Carsten Lutz: could you say what "work" means?
James Hendler: http://www.w3.org/2007/OWL/wiki/OracleOwlPrime
James Hendler: http://www.w3.org/2007/OWL/wiki/OracleOwlPrime#Definition_of_OWLPrime_.28RDFS_3.0.29

Carsten Lutz: I can see two conditions relevant for a fragment: they should provide useful expressivity, and they should be based in some principle that guides the design, such as tractability of subsumption or horness

Carsten Lutz: What would be the guiding design principle of OWL Prime

Alan Ruttenberg: reducibility to implementation in rule bases with large numbers of instances?
James Hendler: -1 to the way Carsten sees defining things, as they don't apply well to the startup/industrial world

Boris Motik: DLP allows to reason about the domain that consists only of the resources that one has in the KB

Boris Motik: no need to generate anonymous individuals

Carsten Lutz: ok, fine with me. Would be nice to learn whether Zhe agrees?!

Boris Motik: OWL Prime is pretty close to DLP

Carsten Lutz: What is the complexity of DLP?

Boris Motik: the guiding principle behind DLP is similar to OWL Prime's goals of using rules for reasoning

Boris Motik: wouldn't be difficult to align DLP and OWL Prime

Boris Motik: OWL prime could be seen as an RDF-oriented version of DLP

Zhe Wu: I agree with Boris

Zhe Wu: should be feasible to map OWL Prime to DLP

Zhe Wu: the criteria in Oracle's mind is to meet the requirements of the users

Zhe Wu: it has to allow an efficient implementation in the context of enterprise applications

Zhe Wu: we like rule sin Oracle because they can be implemented efficiently using DBs

agreed

Peter Patel-Schneider: no
Alan Ruttenberg: +1
Bijan Parsia: Uhm....is this meant to bind us by NDA?
Bijan Parsia: But...if I were to blog it?
Doug Lenat: I'd rather you don't tell us anything proprietary, but other than that, okay.

Peter Patel-Schneider: it puts me in an uncomfortable situation

Doug Lenat: Why not just describe the type (e.g., a large medical center....)
Doug Lenat: Just don't mention the names.
Bijan Parsia: I don't need the names as long as there's concrete details

Alan Ruttenberg: the names should be mentioned but the members of the WG should not mention them publicly

Doug Lenat: Exactly.
Doug Lenat: (Meaning I agree with Peter)
Evan Wallace: I understand Peter's concern!

James Hendler: Tarlick, Radar Networks and Meta Web have given me details about their specifiic needs

Ivan Herman: s/tarlick/garlik/

James Hendler: they think that OWL Prime meets their use cases

James Hendler: these companiens would rather not comment in public

James Hendler: we should take into account the needs of those companies

Alan Ruttenberg: could you say something more about OWL prime meeting their needs?

Alan Ruttenberg: is it about syntax?

Alan Ruttenberg: do they care about completeness?

James Hendler: I cannot answer

Alan Ruttenberg: the problem is that I do not know what they mean.

Ian Horrocks: no need to go any further

James Hendler: they prefer scaling rather than completeness

Alan Ruttenberg: thanks Jim, that helps

Ian Horrocks: we need to agree what to do next with OWL Prime

Ian Horrocks: what are the next steps?

Zhe Wu: we should come up with a set of DLP rules that characterize OWL Prime

Zhe Wu: these rules would define the semantics

Carsten Lutz: Is OWL Prime a fragment then or a conformance level?

Boris Motik: I agree we should come up with a rule set

Carsten Lutz: Or both?

Boris Motik: assume an EL ontology which is DLP

Boris Motik: you turn it into RDF

Boris Motik: and then run the rules

Boris Motik: and obtain the same answers

Jeff Pan: Does Zhe agree that OWL Prime is simply DLP?


Uli Sattler: Jeff, I think OWLPrime is more Full-ish than DLP?!
James Hendler: +1 to Uli


Alan Ruttenberg: my question goes to this

James Hendler: we may have both fragments and conformance levels

James Hendler: there's technical issues that differentiate a full subset from a DLP subset

James Hendler: define a fragment as a subset of a vocabulary and the semantics is the OWL Full semantics

Jeremy Carroll: Please read out the abvove
Jeremy Carroll: yes it's feaisble
James Hendler: but I think there could also be a DL version <scribe assist>
Jeremy Carroll: I would ask HP colleagues, and it's costly
Jeremy Carroll: to get a review

James Hendler: but I think there could also be a DL version

Michael Smith: review what's there now or what is to be specified?
Jeremy Carroll: OWL Prime
Jeremy Carroll: Let's do it
Jeremy Carroll: I'll write an action

ACTION: jeremy to arrange HP review of OWL Prime page

Alan Ruttenberg: I want to meke clear what is the treatment of non-answers

Alan Ruttenberg: I understand that rule systems do not put constraints in the use of classes in the place of instances and vice-versa

Alan Ruttenberg: this does not seem to be in DLP

Uli Sattler: Boris, but this would be a DLP different from the current one

Boris Motik: I think that DLP and OWL prime could really be made equivalent

Boris Motik: OWL Prime is OWL Full like, and we should generate some restrictions

Uli Sattler: Boris, I understand you as volunteering to come up with a unified "rule-based OWL"?
Jeremy Carroll: I am happy to wait
James Hendler: -0 to having to agree to a closed world for OWL Prime - I cannot live with this

Michael Smith: HP should better spend resources later on, when the spec is more advanced

Jeremy Carroll: OK
Michael Smith: ok, that's fine
Bijan Parsia: I'll notes that there isn't a closed world in boris's proposal (it's not non-monotonic)

James Hendler: I don't want to move to closed-world semantics


Uli Sattler: Jim, this is "known domain assumption" rather than "closed domain"

Ian Horrocks: the fragment as specified is decidable

James Hendler: I don't mind if there is a DL version of the fragment, but there should be an OWl Prime Full

James Hendler: I am fine with OWL Prime DL being DLP

James Hendler: as long as there is also a Full version
Uli Sattler: ...but only primitive classes!
Uli Sattler: ...I understand that you can't even have them in the premis


James Hendler: as long as there is also a Full version

Boris Motik: ther eis no CWA in DLP, nor there is a domain closure assumption

Bijan Parsia: +1 to boris
Uli Sattler: alanr, this is what I am reading the OWLPrime description

Boris Motik: what happens is that the use of existentials in the language is limited

James Hendler: I misunderstood what Boris previously said, I am fine with the above - i.e. that the language doesn't allow certain things to happen (expressivity wise)

Ian Horrocks: as I understand conformance level is that it gives you the entailments that should be inferred

Ian Horrocks: it specifies a minimum but not an upper bound

Uli Sattler: I just sent out my conformance level examples per email, and it is like Ian just said

Boris Motik: for certain kinds of queries the answers between DLp and OWLPrime (the rdf version of DLP), the answers should be the same

Boris Motik: I am talking about entailments and non-entailments

Uli Sattler: say again, alanr


James Hendler: most languages I have encountered are defined in terms of rules

James Hendler: but this is not the case in this WG

Sandro Hawke: isnt' a Proof-Theoretic Semantics?
Bijan Parsia: sandro, no
Uli Sattler: Jim, one of the troubles is that these axioms might interact...
Uli Sattler: Saying fullish dlp answers at least set of queries dl dlp. This actually sets a pretty high bar for conformance (or at least I worry that it might) [Scribe assist by Alan Ruttenberg]
Uli Sattler: and then it is difficult to see when we can or should stop
Uli Sattler: alanr, I don't think that anybody is suggesting any conformance level for OWLPrime here
Alan Ruttenberg: not advocating for it, just making sure that we are all on the same page - worried that zhe or hendler might not see consequences immediately and then balk later

James Hendler: we should have an axiomatic semantics

Carsten Lutz: I didn't get that.

Boris Motik: DLs are closer to FOL in that nothing is reified

Boris Motik: so the theory is the ontology

James Hendler: +1 I think

Boris Motik: in the case of OWL Full and OWL Prime, you ahve a set of axioms that reify your theory

Uli Sattler: Boris, I haven't seen an axiomatic semantics for OWL DL...did you?

Boris Motik: the equivalence between DLP and OWL Prime could be established

Boris Motik: for a certain kind of entailments

Doug Lenat: The speakers can edit the minutes.

Boris Motik: those that make sense in DLs

Boris Motik: Uli, the axiomatic semantics of OWL DL can't be given in terms of a set of axioms that is *fixed* for all ontolgoies.
Boris Motik: These axioms in OWL DL would be second-order.
Boris Motik: In OWL DL, the theory is actually the ontology.
Alan Ruttenberg: +1 to Uli - Boris, possible to do that
Uli Sattler: Boris, I understand, but I heard you saying that there was such a semantics
Carsten Lutz: Other fragments for example :-)

Ian Horrocks: next item in the agenda

Uli Sattler: IanH, we have been spending a looong time on one fragment - not on any of the others...

James Hendler: Boris won an award, congrats

Issues

Ian Horrocks: Issue 92

ACTION: bmotik to Send an e-mail to the list with ideas on how to bridge DLP and OWL Prime

Ian Horrocks: resolved

Peter Patel-Schneider: +1

ACTION: bmotik2 to Send an e-mail to the list with ideas on how to bridge DLP and OWL Prime

Rinke Hoekstra: +1

Ian Horrocks: Issue 16

Ian Horrocks: entity annotations? should we have annotations in entity declarations?

Deborah McGuinness: yes - i liked peters 2a proposal
Jeremy Carroll: annotations are not interesting! but they are a duty ....

Peter Patel-Schneider: I proposed to decrease the kinds of annotations allowed

Boris Motik: there is another asymmetry

Boris Motik: annotations of axioms are not axioms themselves

Boris Motik: this causes an asymmetry

Peter Patel-Schneider: this is roughly my 2a proposal

Boris Motik: Matthew suggested to make all annotations axioms

Bijan Parsia: It could be a bit tricky, but it could be simple...would have to work out the details to know which
Rinke Hoekstra: looks like it to me as well
Uli Sattler: http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0106.html

Alan Ruttenberg: do annotations become domain elements?

Bijan Parsia: That's orthoganal I think

Boris Motik: all this is purely syntactic

Peter Patel-Schneider: ... and annotations
Bijan Parsia: No

Boris Motik: we could put annotations on ontologies, entities and axioms

Deborah McGuinness: and i think you can put annotations on annotations right?
Uli Sattler: yes, dlm
Peter Patel-Schneider: ok by me
Doug Lenat: Deb, i think that recursively, by definition, the answer must be yes.
Deborah McGuinness: yes - just making sure for the log. i would use this a lot.

Washington Face to Face

Ian Horrocks: peter requested to ask for sponsorship for OWLEd

IanH and alan: we endorse the request


Peter Patel-Schneider: we would like to get enough money so that we do not need registration fees

Jeremy Carroll: (HP is always willing to host on that basis)
Jeremy Carroll: know

Ian Horrocks: there is an issue concerning observers in the next F2F

Ian Horrocks: we had some issues with observers in manchester

Bijan Parsia: I'm always happy to have observers of any kind
Jeremy Carroll: I feel there should be a cap on number of observers from any one organisation
Bijan Parsia: I'm happy to have observers vote in various circumstantces (e.g., certain sorts of straw poll)

Ivan Herman: we mean observers from member organizations, right?

Jeremy Carroll: Also meeting room arrangement
Bijan Parsia: I don't see any problem with non-member participation at a normal f2f...everything is publically minuted!


Bijan Parsia: +1 to sandro
Alan Ruttenberg: do we have any actual requests for non members yet?
Jeremy Carroll: it raises the bar to making member confidential commetns

Ian Horrocks: it would be possible to have observers from non members, but

Ian Horrocks: they may be asked to leave the room

Bijan Parsia: I do not believe there are issues with non-members

Michael Smith: I will clarify if I know of any such requests
Bijan Parsia: jjc, the bar should be high

Imports Task Force

Alan Ruttenberg: the imports task force meeting is next monday