See also: IRC log
Guus: PROPOSED to accept minutes of 8 May telecon: http://www.w3.org/2007/05/08-swd-minutes.html
Guus: Resolved, minutes accepted
Guus: Next telecon: May 22
Guus: Regarding a F2F meeting -- action continued [http://www.w3.org/2007/04/17-swd-minutes.html#action01]
<TomB:> ACTION: TomB to start questionnaire on date for f2f -- options are weeks of Nov 4 and Nov 11 [recorded in http://www.w3.org/2007/04/17-swd-minutes.html#action01] [CONTINUES]
SKOS Use Cases Working Draft Document
<scribe:> ACTION: Jon to contact Ralph and arrange for publication of SKOS use cases as working draft (May 8 version) [recorded in http://www.w3.org/2007/05/08-swd-minutes.html#action13] [DONE]
Ralph: Thanks to Jon for work on the document
Guus: I would like the editors to send messages to relevant mailing lists asking for feedback, with attention to the new document (like what was done for RDFa);
<scribe> [NEW] ACTION: Daniel to send messages to thesaurus list, semantic web interest, ... requesting feedback on the new SKOS use cases document [recorded in http://www.w3.org/2007/05/15-swd-minutes.html#action01]
Sean: All lists that the questionnaire was sent to...
Daniel: I would like a list of the set of lists to send the message to.
Guus: The request should be sent to the original lists that were requested for use cases.
Daniel: Daniel et al. to get together and come up with composite list to send request to.
Grouping Construct Issues
<scribe> ACTION: Alistair to post links to grouping construct discussion in prep for next week [recorded in http://www.w3.org/2007/05/08-swd-minutes.html#action12] [DONE] (Guus)
Guus: We need some discussion on next two actions: minimal fix and the alternative. The non-ordered approach is structured differently from ordered approach
Alistair: We can divide that into two smaller proposals. Under minimal fix, there is no way to specify that something is unordered explicitly; given an open world assumption, default reasoning is difficult. In the original SKOS Core Guide, there is a discussion of the two different layouts, which may be overly complicated.
Guus: In the minimal fix approach, for the collection case, I would not assume ordering; there is no notion of ordering, for example, with an intersection.
Alistair: As an application writer, if you want to find out that a collection is unordered, you have to examine a graph to see if there isn't something that indicates order. This is essentially default reasoning, which in practice may be a problem. In an open world situation, it would be better to state that a collection is ordered (or not) explicitly.
Guus: The minimal fix solution is Alistair's attempt to address the issue (just the ordering problem) in a minimal way. Another alternative would be to take this proposal as a starting point, and then extend that to provide an explicit feature.
Jon: This would be preferred -- to take this a step further, and provide the explicit feature enabling users to state whether or not a collection is ordered.
Alistair: Referring to the ordered SKOS membership rule -- if you have an ordered collection and member list, you can infer a number of things from queries against the list. This would enable you to ask whether something was a member of a list or not. As is, the minimal fix solution does not make this explicit - there is a level of indirection. With the rule, you can end up with a SKOS ordered collection, you can have both a SKOS memberlist property pointing to a list, and pointers to the members; You could assume if the property is not there, that the list is unordered. Given the minimal fix, we could raise these issues as distinct issues, or provide a strawman that would support fixing them in one go.
Guus: I propose we discuss this next week, and then keep the others as additional issues for the list which we can decide later to either address or drop.
It would be useful, given this, to address the specification of the semantics, and our definition of test cases. We might have to find additional people to work on these issues; Sean might be prepared to take on some of these issues given his experience with the OWL test cases. We need similar infrastructure to support some of this work; write test cases against the specification regarding what an application is expected to do based on the semantics. Sean would be willing to take this on.
<RalphS> [I'm concerned about alternatives such as "if the property is not present, assume..."; that may open non-monotonic issues]
Alistair -- would you be willing to make the minimal fix proposal for next week and raise one or two issues with respect to the potential problems, keeping the rule in, but raising an issue regarding the problems?
Alistair: With regard to test cases, what was the basis for these in OWL?
Sean: There were a number of different test cases, testing different things; test cases include the chunk of RDF and what the results should be.
<Guus> OWL Test Cases http://www.w3.org/TR/owl-test/
If you look at the test cases, you can see examples of various issues, consisting of OWL ontologies and indications of how applications were expected to behave. So here, you would have to describe what the RDF criteria would be, how applications are expected to behave.
Guus: We should take the minimal fix proposal and spend a few minutes on that.
Sean/Alistair: The two examples are inconsistent because they use the SKOS narrowerThan property ... The difficulty with the structure of the proposal - used axiomatic triples, doesn't specify whether it is OWL inconsistent, RDF inconsistent, etc. Even with just an RDF interpreter, we should specify what the behavior should be.
Guus: If you look at OWL test cases, for example, these are the premises, and these are the conclusions expected. This might provide a little more formal form than what is currently in the document. For specification of semantics ... do we think we can handle it ourselves, or do we need additional assistance to work with us on this?
Alistair: We might basically have a number of sections in the semantics, structured similarly to the document currently. If we break up the semantics in that way, the namespace document would be comprised of the axiomatic triples. Then, the issues would be with some of the rules, such as the membership issue.
We might use the way the entailment rules are given informally in the RDF semantics, but I hadn't thought about other sorts of rules yet that might also be pertinent.
Guus: If we can keep this format, which would make it clear to application developers what they need to do, I'm not sure this would be sufficient.
Alistair: This structure follows from the way the RDF semantics is structured, but then if you start to use the OWL vocabulary in the axiomatic triples, then you're no longer building on the RDF semantics. By bringing in semantics from OWL, it would not be clear whether you are bringing in the RDF model-theoretic semantics from OWL or the direct model-theoretic semantics.
Daniel: I thought the goal was to keep it simple ...
Alistair: We could take two approaches, one stating the RDF semantics, the other using some OWL, but if we're using OWL then we have to be clear about how we're using the semantics.
Guus: I suggest that we are explicit about how we use a particular statement from OWL, not duplicating what is in the OWL specification. So if there is a notion of disjoint, we should use what is in OWL not recreate what is in the OWL spec, but could add notes about what it would mean to use a particular construct from OWL. For any formal definitions we should refer to the OWL specification.
Sean: I would like to think abit more about what we're proposing, but would be leery of proposing a kind of mix and match approach. If you're using the OWL vocabulary, shouldn't we be using the OWL semantics?
Guus: Sean might be in the best position to address this. OWL disjoint is a good example to look at - Sean can take a look from the position of an OWL application developer.
In section 2, axiomatic triples: the first four can be part of an RDF schema for SKOS. For the fifth one - disjoint - people have to understand something about OWL to understand how to use this.
Sean: The question is do the rest of the triples fit within the OWL semantics - if they fit in the OWL Full world, then you would have to fit within the RDF semantics (using RDF-compatible semantics for OWL). So there is a user who wants to use SKOS, if this is RDF with a little OWL thrown in, would the file be OWL compliant or RDF compliant?
Guus: We have to point out these cases where you're using SKOS within a DL environment, which may be concerned with property statements, in order to keep within the DL space ... may need to work through this.
<seanb> It's me for the testing!
<scribe>[NEW] ACTION: Sean to look at the test environment supporting the SKOS semantics [recorded in http://www.w3.org/2007/05/15-swd-minutes.html#action05]
<scribe> [NEW] ACTION: Alistair will look at raising the examples from the issues to test cases [recorded in http://www.w3.org/2007/05/15-swd-minutes.html#action06]
<scribe> ACTION: Antoine post link to Relationships between labels discussion in preparation for next week's call [recorded in http://www.w3.org/2007/05/08-swd-minutes.html#action11] [DONE]. See http://lists.w3.org/Archives/Public/public-swd-wg/2007May/0019.html
<scribe> ACTION: Jon and Alistair: Move SKOS issues over from Sandbox to Tracker on an ongoing basis [recorded in http://www.w3.org/2007/05/08-swd-minutes.html#action10] [CONTINUED].
<scribe> ACTION: Guus revise his ISSUE-26 proposal to account for other options [recorded in http://www.w3.org/2007/04/24-swd-minutes.html#action07] [CONTINUED].
<scribe> ACTION: Antoine to make a proposal about SKOS Use Case document [recorded in http://www.w3.org/2007/04/24-swd-minutes.html#action03] [DONE]. See decision, May 8.
Issues admin: Antoine's message
<TomB> [RalphS, are the issues correctly being recorded as "done"?]
<RalphS> scribenick: ralphs
Elisa: I have made some
... I have pulled together a summary of goals and next steps
... I propose to update the wiki to reflect what I have gathered
... and to include Vit's notes
... Tom and I had discussed whether to identify people we might ask to provide input on their experiences w.r.t. best practices
... for version management, some of the things Daniel and his team have done in bio portal would make good examples
... we've not made much progress on identifying best practices here
... but we do know some folk who have articulated policies
... for URI schemes we might talk with Daniel and Chris Welty
... for maintenance policies we have good first-hand experience with SKOS vocabulary maintenance
... Alistair and others have done great work
... goal would be to identify a couple of people who could be good liaisons for each section
Tom: formulate a set of simple questions to ask each group
<dlrubin> FYI--I need to drop off the call now. Thanks everyone.
Tom: it's a big topic, so coming up with a set of questions might be a good first step
<seanb> Sorry -- have to go. Bye!
Tom: we also need to have more people working on this task
Elisa: in addition to myself, Vit
and Siggi both said they were interested in working on this
... some of the folks involved in SKOS said they might contribute
<scribe> ACTION: Elisa to provide outline of work to be done by Apr 17 [recorded in http://www.w3.org/2007/04/24-swd-minutes.html#action11] [CONTINUES]
Elisa: I hope to send summary by next week