SWBPD Informal Face-to-Face

3 Mar 2006, Mandelieu FR


See also: IRC log


Harold Boley, Pantelis Nasikas, Claudio Venicio (Telecom Italia), Fred Boland, Peter F. Patel-Schneider (Free University of Bozen/Bolzano), Brian McBride, Alistair Miles, Ivan Herman, Jeremey Carroll, Libby Miller, Hiroshi Yasuhara (ICeC), Fabien Gandon, Olivier Corby (INRIA), Andreas Harth, Jacek Kopecky, Tim Berners-Lee, Matthew Ellison
Fabien, Alistair, Libby, Andreas


Welcome, Introductions, Agenda planning

Ralph: Discussion on the Agenda for today. Possibility: there are a lot of questions on what happens after the end of April. There was a message with early draft of new charter proposal for Semantic Web activity. Can talk about that.

Jeremy: main question is for each group which organisation is willing to participate.

Ralph: The plan for the next Tech Plenary week is to combine it with the advisory groupe november 2007 (18 months from now)

Peter: it is a problem because it (possibly) conflicts with ISWC
... someone should raise the problem.
... the plan is to have the advisory meeting on thursday morning and thus WG would work from thursday afternoon to saturday morning


(next plenary meeting schedule)

Ralph: any other agenda item?

Peter: can we really talk about re-chartering activity?

Libby: need a quick slot to talk about how to fix the page of ADTF.

Ralph: Alistaire do you want to talk about SKOS / Voc Management?

Alistaire: there's a couple of issues we could talk about (OWL formalisation)
... cookbook

Jeremy: I will be only working in the HTML and RDF.


Libby: about ADTF

<libby> http://esw.w3.org/topic/SemanticWebDOAPBulletinBoard

Libby: the page has to be reviewed/updated/improved. We are not really evaluating this.
... you have to have a DOAP description, this could be a bit more flexible
... this list is more useful than anything I have done with the list
... this list is useful in itself and we ought to invite people to put things in it even if they don't have a DOAP desc. anyway

<scribe> ACTION: Libby update the taskforce homepage. [recorded in http://www.w3.org/2006/03/03-swbp-irc]

Ivan: a lot of links give 404

Libby: we should find a a way to review / fix it.

Ralph: Data got lost in an accident and we were hopping to fix it/recover it.
... what happens next?

Libby: happy to continue to invite people to add more and more stuff on the page.
... it makes sense to continue this in the interest group and enable the Outreach group to use it.

OWL Futures

Ralph: move to SKOS and OWL

Peter: OWL 1.1 updates OWL DL
... no major changes, a few fixes (see homepage at manchester)


scribe: there is a mailing list world-accessible.
... committment of major DL-reasoner developpers to implement 1.1
... changes: (a) qualified cardinality restrictions (at least 3 children who are doctors)

Ralph: OEP intends to write a note on that

Peter: (b) certain kinds of rules (the friends of my siblings are my friends)

Brian: my enemies enemies are my friends

Peter: not decidable
... punning for every class like person there is also an individual that just shares rhe same URIs just syntactic sugar no reasonning.

Fabien: it could be used in the usecase like books about Lions

Peter: OWL 1.1 cannot have an RDF extension semantics.

Ivan: you cannot map it on proper RDF

Jeremy: if you say 2 classes are identical you are in OWL ffull

Peter: the OWL 1.1 semantics is weaker than RDF

Ivan: it is not an insignificant cost.

Jeremy: huge cost.

Peter: anybody who is willing to join the WG is welcome.
... OWL 1.1 is both implicitly and explicitly is not an XG because it does not comply with W3C view
... the delay for an XG is also a problem
... the target is to have everything in the public domain
... academic group at the moment
... time frame: started in november 2005, design deadline was January, target some implementation befor WWWC 06
... RACER might be pretty close, CEREBRA not sure
... by may the work should be done

Ivan: talk to Dev Day track at WWWC

Jeremy: deadline for Dev track two starting monday 6th March

Peter: probable plan would be a member submission (manchester, bolsana, mindswap, racer systems, cerebra) would include all the technical stuff
... after that we don't know. Many of the people who put the requirements together are not W3C members more from bio informatics groups

Ivan: that's changing with the Healthcare and life-sciences WG

Brian: adding more functionality to OWL. There is a big gap between RDF/S and OWL there could be a sweet spot iin between.
... what about an effort on the simplification of OWL.

Peter: some people have proposed (by usage) simplificatio of OWL
... not aware of atempts to formalize that.

Harold: which functionalities could be put in OWL Lite?

Peter: punning coudl go. I have no idea really cause I have no idea what OWL Lite should be.
... the design of OWL1.1 is a functional syntax
... OWL 1.1 is driven by implementors who know what can be done in the DL framework.
... the thing that has caused most problems in OWL 1.1 is annotation
... comments are supposed to pass comments all along their processes.
... comments are not associated with resources

Ivan: comments are not map to triples at all.

Peter: exactly

Jeremy: what was defective about the annotation properties in OWL 1.0

Peter: I thought that they wanted annotations on resources and with punning you could add pretty much anything you want.
... but they want to be able for instance to preserve the order e.g. this is a comment about why this label is the fouth label.
... we keep the comments as long as possible

Ralph: could we use this to put processing instructions?

Peter: no connection between comments and processing

Harold: what about instances? different ways to represent instances?

Peter: concept definition, concept inclusion, instance definition: sames as OWL 1.0

<RalphS> Ralph: "... (gag) ... processing instructions ..."


<aliman> http://isegserv.itd.rl.ac.uk/blog/

Alistair: on SKOS, the OWL futurs is relevant to SKOS
... first (small reason) : trees of topics are used to retreive all the results that have narrower topics

<RalphS> http://www.w3.org/TR/swbp-skos-core-guide/#secindexing

Alistair: looking at SKOS Core Guide / section subject indexing
... associating a log entry with a subject
... currently expressed using a rule (but no standard syntax); here we could use OWL 1.1 to do that

<RalphS> Alistair: "Subject Generality Rule" could be expressed with OWL 1.1 role inclusion

<RalphS> http://www.w3.org/TR/swbp-skos-core-spec/

Alistair: second (big reason): we have an issue in current draft
... number a situation we want to import SKOS in OWL
... SKOS being in RDFS we cannot do it.
... some of the annotation properties are sub properties of other properties which send us out of DL

Peter: in OWL 1.1 you would use the punning instead of the annotation properties
... you could make rdfs:label a classic property if you say its value is a litteral

Alistaire: so you would get rid of annotation properties and use punning to get the same effect ?

Peter: exactly

Ralph: the spec says the range is a litteral

<RalphS> "The rdfs:range of rdfs:label is rdfs:Literal"

Peter: then it's fine.

<RalphS> -- http://www.w3.org/TR/rdf-schema/#ch_label

Alistair: one option was to have two versions RDF/S vs. OWL but that encourages the divergence between the two langages.
... another option is to have a shared core and extensions
... do you see annotations properties should disappear from tools such as Protégé?

Peter: they should disappear.
... I can't speak for the Protégé people, but there was an idea to move the SWOOP thing in
... Manchester is highly motivated to have something working with OWL 1.1

Ivan: I can understand the reason for the academic demande, from W3C point of view I see a huge demande to get SKOS out.
... we shouldn't wait to much.

Alistair: I propose quick fixes to these problems
... but we should anticipate futur evolutions

Ralph: we must avoid duplication when the OWL 1.1 arrives.

Ivan: what about the rule pb?

Alistair: we would leave it as a rule.

Ralph: when the RIF specs comes arround we adjust.

TBL: interested in best pratices on how to handle/publish different flavours (RDFS, DL, Full etc.)

Alistair: indeed this is a generic problem.

Jeremy: how could we address interop issues between OWL DL and OWL Full

Alistair: Tim presented different options could we get the opinion of peoples in the room.

TBL: one option is to publish a maximum and it is up to the reasonner to define what it can do.

Peter: I don't agree with this way of descriing the pb. If there is something that does not fit in my view of the world I would like a transformation to exist out there to make it compatible with my reasonner.

Jeremy: we could have mapping rules to know what we should discard

Ralph: I am under pressure to get the SKOS rec out. I beleive it is not possible before the end of this WG.
... we have to push back put we should stage the SKOS rec. and not wait the OWL 1.1

Alistair: should we release an OWL full version and let the reasonner handle the transformation?

Peter: that would be the minimal satisfying solution.

<JacekK> JacekK: I thought the Semantic Web is based on the principle that what is unknown is ignored, so an OWL-full ontology fed to an OWL-DL reasoner should be processed as OWL-DL, ignoring whatever is outside DL limits

TBL: +1 with Jacek

Jeremy: it is a very hard issue, we may have the energy to reopen it
... but both side had a case and it is going to be difficult.

Ralph: so we should let it bake for a while, in some application for a while.

Jeremy: we need to bring everyone at a table to reopen it and discuss it deeply.

Harold: everything is optional, sometime only a whole set of properties makes sense but not a sub set.

TBL: would that mean that with a sub set we would derive false statements?

Peter: not sure but possible case.
... if I have a collection of RDF triples if I remove certain of them (e.g. the thing that cause a negation) you might have a problem.

<scribe> ACTION: Peter to informally determine weither removing triples could flip a boolean in RDF and send email on that [recorded in http://www.w3.org/2006/03/03-swbp-irc]

Jeremy: close to an action in WebOnt on the layering of OWL Full on OWL DL
... I have worked on that for 3 or 3 months and couldn't answer that.

Harold: cardinality restriction contains such a negation

Peter: open world: it does not become false but "not true"
... suppose we could separate all the axioms then it would be a monotic action to remove the triples (Scribe is lost here)
... in OWL this might not be the case

Alistair: I wanted some advice on what to do in the short term.
... what I am hearing add OWL typing, not worry about in OWL Full but explain how to remove them

Peter: yes and explain the link between them and which ones to remove together

Harold: broader / narrower is the same than subClass?

Alistair: No.
... in a thesaurus it could be part of for instance. What you are saying is that a doc about a node is also about another node.

Harold: so from super class you could imply broader?

Alistair: pandora box
... needs a long discussion. Long messages on the mailing list. I see that as less vital for first version of the specs.

Harold: connection between SKOS and RDFS is also interesting.

Alistair: it needs to be explore, same for SKOS and OWL.
... these are additional issues should be dealt separatetly, not in the short term.

Ralph: new people introduction

Adreas: DERI Galway (RDF storage, data integration, etc.) in SWBP

Jacek: Deri Innsbruck

Matthew Ellison: STC, looking at the way W3C works, getting familliar with subjects

Tim Berners-Lee: (do I really need to present him ???)

Ralph: should we add SKOS/RDFS and SKOS/OWL to agenda?

Vocabulary Publishing Cookbook

<RalphS> http://www.w3.org/2001/sw/BestPractices/VM/http-examples/2006-01-18/

Best Practice Recipes for Publishing RDF Vocabularies

Ralph: copy and paste recipes to configure a server to publish a voc.
... we had to pick a sevrer, these recipes are for Apache.
... they contain the extracts of configuration files ready to be pasted
... examples of hash ns and slash ns

TBL: something came up in the tabulating example.
... in practice there is a issue with the slash ns you have to make an 303 check for each property
... a lot of HTTP activity for a document.

Ralph: you mentioned with a hash ns once you fetched it you know everythinh you need to know.
... with wordnet you don't want to know everything you need to know.
... should be published with in the next days.

<RalphS> Fabien: is it a use case for SPARQL DESCRIBE to want to know more about a property?

Alistair: I imagined a situation where you could use a SPARQL service, but how do you know you're dealing with a SPARQL service.
... there is a tool to partition your ontology to identify the bit you might want to send/reteive.

Ralph: the WG should leave it as a working draft and the voc. management WG will take over.

TBL: good job at how to deal with 303, should get this work out, so go ahead and include possible improvments as inputs for the voc management WG.

Alistair: the sixth recipes would be the Wordnet case.

TBL: isn't it example 2

<libby> al: the 6th is a placeholder

<libby> ...for wordnet-like things

TBL: it would be nice to have two or three examples.

Ralph: fixing Dubling Core to do 303 is an engineering challenge.

Alistair: relationship between this and RDDL.
... this doc tells you how to get the HTML, the RDF

<RalphS> Resource Directory Description Language (RDDL)

<aharth> hi danbri

Alisatair: you could imagine a situation where so writes a schema but uses RDF/A

... a TAG best pratice says that it is useful to have a human readable version
... I would like to have the RDF/OWL and have a stylesheet attached for the human readable version (using processing instruction)
... the idea is to get the RDF / OWL out first

Ivan: a bug is mozilla is that it does not understand blabla+xml

Ralph: it works my Mozilla although it has some issues with the refresh

Ivan: I never found any way to do it in Firefox

Ralph: other topics on cookbook?
... this doc should be a tech report next week.
... we adjourn to lunch returning at 2pm for Activity Rechartering.
... other topics?

Harold: negation.
... also RIF vs. OWL, Horn logic that does not have negation.

AListair: we could discuss SKOS and OWL relation too.

Ralph: 30min on Rechartering, 30 min on Negation & RIF, 30 min on SKOS & OWL

<danbri> re "Alisatair: you could imagine a situation where so writes a schema but uses RDF/A"

<danbri> ...prime candidate for that is the xhtml2 link types vocab

Activity Rechartering


<RalphS> http://lists.w3.org/Archives/Member/w3c-ac-members/2006JanMar/0038.html (member-only)

This message went out last week, pointing to draft activity proposal. This week Eric M has been collecting input on this draft, will collect input for at least another week then produce final draft.

The existing groups under the current charter i.e. DAWG and RIF continue, no changes there.

Also semantic web for life sciences and semweb coordination groups are extended, but no real changes to the charters.

(btw Ralph is talking above)

Ralph: propose to split best practices into RDF extraction WG, vocabulary management WG, and education and outreach ...

GRDDL is expected as recommendation under RDEX - this a new piece of work we are proposing.

SWVM is proposing some things from best practices as recommendations, especially SKOS.

ralph: (continued) ... two things are noted as not addressed in the current activity proposal: (1) is OEP work, there are questions as to whether this could be combined with SWVM or should have separate WG (2) RDF/A does not have a home yet.
... Applications and demos will by picked up by SWIG, MM members are drafting an XG proposal, OEP is an open question, PORT goes into SWVM, RDFTM may become an XG, SE may become an XG, VM goes into SWVM, WordNet is not addressed in currect activity proposal, and won't be unless strong request from the members, may be another XG, XSCH work is done>

Peter: best practices was good becomes formation of task forces was open, bad thing about formation of XG is that process is closed. I.e. W3C does not talk about formation of XGs and WGs as openly as the Best Practices talks about formation of task forces. I consider that between a shame and an embarrassment.

ralph: I agree with the fact, but it is importatn to members to know the scope of a WG before hand, esp. wrt patent and policy. E.g. we had discussions in best practices whether we could do rec track work, and it was important to have made that statement up front.
... If we had declared a document to be on the rec track, that would have invoked aspects of patent policy that we could otherwise ignore.

peter: better to make the formation of semweb activity XGs on public lists, so discussions can be in public.

ralph: opportuniy to continue best practice work exists as XG or as TF in SWIG, and SWIG TFs are proposed and discussed in public. Although SWIG has not done as much formal publication as an IG might, it is within its charter to do so.
... Any of the working drafts or notes that best practices produced could have been published by SWIG.

Peter: problem with proposed structure is stealth XGs within semweb activity.

<ivan> ScribeNick: aliman_scribe

ralph: I share concern you have about XGs. We the W3C staff don't necessarily know of discussion to propose an XG until a proposal is submitted. Members are under no obligation to disclose plans for XGs. I am attempting to set precedent with MM XG that they have all conversations in public.

peter: we would like all of substantive recorded conversations to be in public.

ralph: the AC reminded the W3C staff that they wanted to be informed during drafting of activity proposals and WG charters. We're not sure which formum is appropriate for this type of discussionl.

peter: problem with chartereing TF under SWIG is that SWIG is not as active or controlled as BP was.

ralph: is this good bad or neutral?

peter: good because can propose things outside charter of BP, bad in the sense that flaky things could happen. I.e. hard to forbid people from forming a SWIG TF.

ralph: this is partly by design.

peter: problem is not the formaition of these TFs, the problem is that these TFs may pollute the mesage, esp. if there is not the same level of reveiew that the BP has been trying to do.

ralph: when people approach us in hallways to ask for a mailing list on blah, we can point them at the IG.
... N.B. the term Task Force does not appear in the W3C process. A TF is a proper subset of a WG that does something, but the WG as a whole is repsonsible for the deliverable.

peter: in the case of an IG?

<danbri> (SWIG TF's happen via SWCG, so there is a level of filtering and discussion, albeit not public)

ralph: then it would be an interest group note.

<danbri> (but agree it's not WG-like (by design))

ralph: we conduct concess within a WG whether there is consensus, harder to achieve in an IG.

<danbri> (especially an IG with such a loose notion of membership; some other W3C IGs have enumerated memberships)

ralph: e.g. the SWLS has much more formal structure, so can measure consensus. there is nothing charterwise that distinguishes SWIG and SWLS.
... my reaction to danrbi's comment on formation of SWIG TFs - this is way SWIG chair has chosen to solicit feedback on formation of task forces, doesn't necessairly satisfy peter.

peter: ideal way to get a note out a TF in SWIG would be to get general approval by the IG, with a review by the CG, with expectation that review is proforma.

<danbri> The http://www.w3.org/2001/sw/hcls/charter.html#membership and SWIG charters do differ significantly. SWLS counts as its formal participants only W3C Member reps and Invited Experts; SWIG is totally public (for better or worse)

ralph: at a general level this is the funtction of the CG, it is where all the charis of the activity get together and coordinate, it also has former chairs and others, whoever activity lead considers helpful.

peter: this could create more work for CG.

ralph: what makes BP more open is the mailing list, altho its only responsible for responding to itself.

peter: SWIG will never have a participants list? Only way to guage consensus is to look at the mailing list?

<danbri> (I made point on CG list recently that the SWBP served a coordination function somewhat like the CG itself)

ralph: yes. we would like to find a cochair for SWIG, that may materially change some of these process issues but we don't expect it to do so.

<pfps> Yes, I noticed Dan's comment, which fed into my comments.

<danbri> ...I'm interested in idea of SWIG having an enumerated membership

<danbri> ...but the mailing list membership is not public info

ralph: the mail message to the AC re SWAP3 says w3c-ac-forum and or w3t-semweb-review are places to send feedback.
... this is 3 months late.

<danbri> ...it isn't clear that signing up to a W3C list entitles W3C to advertise that fact-- though i've not studied the privacy policy

Ivan: I have seen no reply yet.

ralph: we have had some feedback already.
... fedback on OEP and RDF/A work is particularly important, as we intend to continue this work.

ivan: was does HTMLWG say?

ralph: we intend that ownership of RDF/A spec be entirely responsibility of HTMLWG.
... in semweb activity we continue working on RDF.A primer, but formal spec of RDF/A is mixed thru HTML spec. So RDF primer is really for semwebbers.

ivan: there were issues e.g. the original design was to have RDF/A as part of XHMLT 2 ... I have heard voices from XHTMLWG that if it goes another way ... if it becomes a module you can use with XHMLT 1 that would be OK.

ralph: original paper by mark birbek makes RDF/A a module of XHTML 2. It would take a different RDF/A design to make RDF/A a module of XHTML 1 ... certain pieces of RDF/A would not work with XHTML 1.

ivan: based on what you say, any design of the issue i.e. whehter RFDF/A is bound to HTMLWG ...

ralph: decision falls to the CG on what charters they approve and not.

ivan: XHTMLWG is also rechartered, are these two recharterings are synchronised?

ralph: we're not hearing strong pusshback from members to do RDF/A module for XHTML 1.

ivan: steven has mildened his stand on this.

ralph: HTML charter is also behind.

ivan: the SWVM would include SKOS, include work done by Thomas B ?

ralph: SWVM charter lists three deliverables (1) SKOS (2) the cookbook (3) best practices for publishing schemas and ontologies, versioning, managing namespaces.

ivan: there were also voices about defining core ontologies.

aliman: third SWVM deliverable may be better broken into multiple docs.

ralph: no problem with that.

peter: couple of other issues ... RDEX charter ... targetted towards current funcitionality of GRDDL ... supposed to output RDF/XML ...

ralph: feedback thru AC rep.

aliman: put D2RQ or something similar in RDEX.

ivan: reaction of people in that group, with buzz around microformats, high priority to have GRDDL finalised and out very quickly, will push back on anything that slows that down.

peter: doesn't preclude followon transforming into other formats.

ralph: in the email to the AC there is mentioned the issue of relaitonal databases as RDF.

peter: base design for GRDDL where input is HTML, output is RDF/XML?

ivan: yes, this is the base intention.

peter: this is desirable. It would be a shame if microformats moved in directions that GRDDL could not handle.

<libby> brian: what would like to see standardised there

<libby> al: d2rq vocab as a note - probbaly enough

<libby> ...people would build tools on top of the mapping vocabulary - build a mapping between an ontology and a relational model

<libby> brian: agree that it would get lots of data onto the semweb....not positive personally that it's needed - tools there alreday, working in different ways - is there a need for interoperability there?

<libby> ralph: many of us feel there is a need for some technology to point people at - like apache cookbook - you don;t 'have' to, but can point people at it

<libby> brian: that would be v vaaluable, sound slike an XG

<libby> al: I think d2rq works on sesame and jena - kowari is diferent I think. beyond XG?

<libby> [missed some from peter]

<pfps> [not important]

<libby> al: wants a tool that I can go click-click-click and have it spit out some stuff

<libby> thanks peter

<libby> ralph: possibly a WG isn;t the best way

peter: navel gazing activity?
... i notice various philosophical issues that are being solved in different ways ... e.g the RDF/XML stranglhold is being broken in different ways, not sure compatible ...
... e.g. SPARQL has a different languag, RIF ...

ralph: it's the role of CG to apy attention to those things

ivan: we had discussion of giving formal status to turtle as a serialisation

peter: i think bigger than just tutle and n3

ralph: role of CG to do this.


Al: a number of different docs and usecases have suggested that we need to understand relationship between SKOS and OWL

-looking at classes and values note

<RalphS> http://www.w3.org/TR/swbp-classes-as-values/

Al: classes as values of properties is owl full (or 1.1)
... class as values notes was written as design patters to get around this problem

<pfps> navel gazing == philosophical foundations of the SW [repeat because I was off-line]

1: point directly from a book to a classs = out of owl-dl

2. I have a book which is about a thing that is a lion - DL safe but doesn't really capture idea trying to say - about lions

3. basically - to express the idea of lions and of african liions, you create intermediate nodes that express lions as a subject or topic of discussion - my book is about this intermediate node and seealso this classs in the ontology

scribe: using rdfs:seealso

(1,2,3 were al speaking btw)

al: part of (3) is what skos is trying to do - general relationships without logical rtelatiiopnships [missed some]
... thesaurus describes the books but there is also a relationship between skos and the ontology
... so this is the issue - how is this relationship expressed.
... another usecase: Libby and Al are both using flickr but we have a different flickr tag for danbri. Flickr tags can be expressed in skos
... somewhere also issomethig that says danbri is a foaf;Person
... danbri came up with a specific property for this connection - 'this means the same sort of thing'
... search for the word 'denotes' on the esw thesaurus mailing list - long dicsussion

<RalphS> Alistair: search for 'denotes' in http://lists.w3.org/Archives/Public/public-esw-thes/

al: Another useacse: I have a thesauarus and generate a skos version of it
... I could put more meaning in my thesuarus - disambiguate some of the broader / narrower using an ontology

[missed some]

ivan: 'is' or 'bound to' ?

Al: 'is'

<danbri> 'it' and 'as'

ivan: why can;t I use typing to directly link to the class that it is

al; no this is the case where there's *no* ontology - only skos

<RalphS> ScribeNick: libby-scribe

scribe: question is, is it ok for them to say this thing - which was a concept in skos - is really a class - and this relationship between 2 nodes - previosuly 'broader' is actually a subclass relationsip
... is this ok or not
... previous approach to keep ontology and thesaurus seprate; this is a different approach - beef up thesuarus

<danbri> (i'd say not; thesauri are too messy... 'broader' gets used for lots of things, only some map to subclassof)

fabian: part of the thesaurus or all of it

al: all of it
... is the usecase; lots of thesaurus people have heard of ontologies, know they can do something with reasoners - want to do this
... 2 design patterns - ontology and thesurus seprate, or somehow the same thing?

harold: related to punning

ivan: the big selling point is the nutcracker - clear, clean, people understand it easily, even if they don;t understand owla nd don;t want to
... explicitly bringing these things into skos makes for a hybrid sledgehammer / nutcracker
... of course we could not 'ban' it - but let's not give them the rope to ahng themselves
... my take is to keep separte

al: I consider these usecases to be edgecases for skos, not essential

ivan: for the ontologies, adding the bridge might help

al: rdfs:seealsdo from classes-as-values doesn;t capture this meaning at all of the relationship
... so we should coin a property that does capture the semantics

ivan: or we use rdf:type

al: doesn;t work - ...you'd have things that have type class, type property, and (type) instance

ivan: ok yes, I take it back
... property then
... first one - the property is ok - 2nd one hard to sell

al: argument for the 2nd merging one is from thesauri experts
... why do we have both

ivan: tell them - feel free to use the sledgehammer

al: I could happily live with that

fabian: interoperability? using same uri...we are talking about exacvtly the same thing so use the same uri for it

ivan: semweb is very precise, but most people have a usecase for much less accuracy - they still want to describe it - these are different audiences
... they don;t care - somethign elsy and clean to do

fabian: totally ok with people who want to use kjust skos - you are just doing the same thing as befgore - but what if you want to link it - then should you not use the 'best machanism' - it makes sense to me to use the same uri or sameas

ivan: extra property in skos that makes the connection clear - this is clean, and then you can have systems that can use both
... uri for the two - smart people can, and we cannot stop them - but too much rope

fabian: we end up with seealso, describedby, etc etc - more and more confusing...this proposal same uri is much cleaner

al: this is just like the conversation on the list
...: -)

al: I can see more points of view. would like to add soe more details
... agaianst owl:sameas type thing: concept in one thesaurus and one in another; one has lables and definitions and comments - suppose they sameas the 2 concepts? then the nodes get merged; the labels, comments and definitions on each one get mixed up
... also they may change
... allowing sameas between concepts is problematic. so we say 'never use owl:sameas'
... there is a skos mapping vocab though currently not deeveloped. in there is 'exact match'
... think this is a very strong argument against punning
...also: harold saaid earlier what is the relationshiop between subclassof and broader than. if they are separate worls - no relationship between these exists. if we combine the worlds then they may have some kind of realtionship, maybe subproperty
... thbe main issue was - what the hell do we call the new property?
... the 'amstaerdam proposal': skos:as and skos:it
... from skos to ontology: 'it'; form ontology to skos - 'as'
... not very m,emorable

ivan: prefer 'is' and 'as'

al: would make a good tshirt though

harold: [someone] did a taxonopmy for chemistry at twente

<RalphS> Harold: Paul van der Vet: subject vs. class

harold: you have different refinements of broader 'isa' vs 'classof'

al: thought this were the same

fabian: no...

harold [missed some]

al: parphrasing: idea of a broader relationship can be refined intoi several subrelationships - part of, subclass of +2 more
... this borader property could be used to link up properties in both skos and ontology worlds
... partial order relationship - don't know if it's part or subclass - can be hard to see the distinction sometimeszs; not always important but is for logicians
... tree circularity is a proplem in thesauri so practically broader should not be reflexive; but if [...] use it with ontologies it has to be reflexive
... thesauri might end up with cycles in thesauri - this has a meaning in subclass but not clear what it means at all in thesauri
... would really like links to relevant ontologies material

[place to look it in rdf core]

fabian: new topic. some people using skos as a half-baked ontology; others use it for capturing labels
... we have a user of skos for organising elearning
... it's very useful to them. We have a stylesheet that transforms it to owl lite
... is skos like the label value of a class - orr is it the class itself
... denotes is symptomatic of that

al: would really appreciate a short description of that problem
... peter, do you ahve any thoughts here?

peter: I'm sympathetic to the idea of this effort.
... has some problems with intermediary notes
... suggests 'punning'
... individuals show up automagically with punning

<RalphS> Peter: punning allows you to use the same name for a Class and an instance

peter: making superproperties of subclass of can cause problems with craeting reasoners
... if you want the reasonser, you need to distill out the skos world

[not sure I got much of that sorry]

harold: rdf is metadata - shoudl be used for documents - so you end up starting talking about both classes and documents - ambiguity

al: as I understood the first pattern, you make a distinction between a book, a class of things in teh world and an idea - books, skos, owl
... separating out classes, individuals and properties form thoughts
... not mixing books and classes

harold: dublin core is metadata, only later became part of logic

[discussion about rdf as KR language vs metadtaa language]

ralph: recap of things left to talk about: RIF / OWL ; Negation


[Peter's discussion refers to a diagram that he drew and updated interactively]

pfps: world according to rif and owl

<RalphS> pfps: datalog is function-free

rdf is the basis, then you can extend that to OWL (DL) -> SWRL -> FOL, or Datalog -> Horn

rif tries to interchange between SWRL, Horn, production rules systems

production rules often called ops rules (ops was the first system)

<RalphS> PFPS: production rules aka "OPS rules" can have side effects, such as counting

swrl and horn view the world at least somehow in a similar way, production rules systems have a different worldview

<RalphS> pfps: Flogic totally includes RDF

fabien: is it possible to encode RDFS in Flogic?

pfps: yes, but flogic has second-order syntax but first order semantics

<RalphS> pfps: to be second-order you must be able to say something about arbitrary sets, not just a set you can compute

harold: higher order syntax is hilog, but often this is subsumed under the term flogic

pfps: truth predicates can lead to trouble even in fol, certainly in flogic
... sparql is not written in RDF

harold: syntax is more like sql

ivan: more combination between sql and turtle

pfps: sparql-queries cannot be written as triples
... swrl can be written in triples via a mapping
... if you took flogic or hilog and like the idea that classes properties and individuals live in the same area then flogic is a good way to go
... trouble is that logical symbols (e.g. quantifiers) in RDF can be uris, and you can change the meaning of those uris

ivan: all of the languages can use rich datatypes?

pfps: yes, they can have built-in datatypes, in flogic you can even construct the datatypes (e.g. lists...), for full treatment you'd need to second-order logic

harold: both flogic and rdf have minsky's notion of frames, and they are object-centric

ralph: what's the best practices impact?

pfps: what should the rif syntax look like?

harold: charter says xml syntax

ivan: i want to understand why xml syntax is a "must"
... xml syntax is not meant for human consumption

pfps: whether everything is uris, what about qnames...

<RalphS> "The primary normative syntax of the language must be an XML syntax. Users are expected to work with tools or rule languages which are transformed to and from this format."

<RalphS> -- http://www.w3.org/2005/rules/wg/charter#xml-syntax

<RalphS> andreas: scribing offline

<RalphS> ScribeNick: aharth_scribe

pfps: problem with the production rule systems which are procedural

harold: maybe production rules are not part of the sw, but in the area of sw services

pfps: what kind of semantics are allowed.

ivan: semantics of datalog and swrl are different. how to bridge that semantics?

harold: if you don't have negation, you can use the same semantics

pfps: in the presence of negation, you notice the difference

ivan: i can create a model theoretic semantics up to swrl, and if there is no negation the semantics is the same?

pfps: yes, they behave the same, with negation it's different

ivan: how do you find an intersection?

harold: small subset of production rules which only have assert then there is an intersection

<RalphS> ScribeNick: aharth_scribe

pfps: some of the rule vendors want to model their rules with certain characteristics

fabien: a lot of work of ilog is around workflows

pfps: manual steps are involved in workflows
... another example are rules which are in cisco routers and such
... sometimes the use of the term "business rules" is ambigious

ivan: the discussion should be written down that people from other schools understand

pfps: rif will be a formal language with some sort of meaning, then you'd be able to write a native rif interpreter
... another issue is that an antecedents look like sparql queries
... sparql functions should be side-effect free

what's the relationship between rif and sparql

horn and datalog semantics match basic graph pattern queries (sparql) pretty closely

fabien: cannot use a function in the construct part?

pfps: sparql is declarative in a fairly strong sense

ivan: where does cwm fit?

harold: cwm is esentially a production rule system
... also has extra-logical things such as retract

pfps: where does n3 (minus cwm) fit in, n3 as a non-active language?

harold: context is in n3 and also in sparql

pfps: unlike most query languages sparql doesn't have closure

alistair: how general rules fit in, especially how a rule can be consequence of another rule?

pfps: some extensions allow rules with rules as consequence
... don't create, they are formulae

(a & b) -> c and a ->(b->c) are queivalent

harold: production rule systems can create new rules

pfps: procedural implementations of this work fine

harold: scope is needed for negation (aka context)

aharth: context has been removed from n3 site

harold: context has been investigated in sintek & decker's triple paper

pfps: what about subcontext and contexts?

ralph: you can name a context and refer to that name

aharth: remote contexts where you can send queries and retrieve triples via http


ralph: how much time to allocate to the next discussion?

harold: mainly provoke a discussion about negations, closed world and open world in the semantic web

harold: trouble when bringing together open world assumption of owl and closed world assumption of rules
... scoped negation as failure might be an option
... owl as open world is a nice match with the web at large
... sometimes you need a closed world: all official w3c reports are at a given site

fabien: have the same type of scenario: closed world assumption about a namespace, if somebody is not inside a namespace, he's not a member of the organization

<RalphS> ralph: that's what rdf:Collection does

ivan: a basic dl reasoner doesn't give you that answer?

fabien: yes, that's an extension
... extension closes a set

<RalphS> Andreas: if you have a notion of a set of triples at at URI then you can use negation as failure

<libby> [whiteboard photo: http://www.flickr.com/photos/nicecupoftea/107227941/]

<RalphS> Andreas: how difficult would it be to add this sort of thing to OWL-DL reasoners?

<RalphS> pfps: certain kinds of closures are quite easy, others are extraordinarily difficult

<RalphS> ... many of the closures that you come across in practice can turn into an OWL-DL statement

<RalphS> ... e.g. the class of American Citizens ...

harold: if the system is decidable (such as datalog), a dl reasoner can call the rule system and use results in further dl reasoning
... there is a problem related to reducing n-ary relations to binary ones
... vague analogy to closed world reasoning with scopes
... predicate-centric would be another way of closed-world assumption

either draw a random line around a sub-knowledge base, or you have a global assumption that e.g. inriaemployee predicate is closed

<RalphS> Andreas: but if I make a statement using this predicate at some other URI ...

<RalphS> Harold: there must be an authoritative Collection of links somewhere

fabien: their namespace approach is used similar to tracking the source

harold: you'd need to have a central point for coordination
... the issue of negation in the formalisms is still open

ivan: there are people working on that, people are interested in probabilistic or fuzzy logic
... black and white view of semantic web doesn't work for the life sciences people

pfps: sure, somebody's got to do work to make this story go well

<RalphS> pfps: everything on my diagram referred to discrete-valued logics

ivan: some people met in galway and thinking about an xg about that topic

pfps: how infinitary valued logic can work on the semantic web?

<RalphS> pfps: xxxx at Stanford has done some work on infinite-valued logics

ivan: wasn't fuzzy logic not part of the charter?

harold: yes, as a possible extension to part 2

pfps: setting up a workshop where users who want these things would be interesting

ivan: ken laskey had a workshop at iswc in galway


Summary of Action Items

[NEW] ACTION: Libby update the taskforce homepage. [recorded in http://www.w3.org/2006/03/03-swbp-irc]
[NEW] ACTION: Peter to informally determine weither removing triples could flip a boolean in RDF and send email on that [recorded in http://www.w3.org/2006/03/03-swbp-irc]

[End of minutes]


$Log: 03-swbp-minutes.html,v $
Revision 1.2  2006/03/06 21:14:46  swick
Cleanup for publication
Revision 1.1  2006/03/06 17:21:04  swick
First rough draft

$Date: 2006/03/06 21:14:46 $