W3C

- DRAFT -

Provenance Working Group Teleconference

27 Oct 2011

Agenda

See also: IRC log

Attendees

Present
Curt_Tilmes, Yogesh_Simmhan, Sandro, dgarijo, smiles, Satya_Sahoo, SamCoppens, +1.518.633.aaaa, Yolanda, Luc, stain, pgroth, [IPcaller]
Regrets
James_Cheney, Christine_Runnegar
Chair
Paul Groth
Scribe
Curt_Tilmes, Curt

Contents


<trackbot> Date: 27 October 2011

<pgroth> can I get a scribe?

<Curt> I'll do it

<pgroth> thanks curt

<pgroth> Scribe: Curt_Tilmes

<pgroth> Scribe: Curt

Admin

<pgroth> http://www.w3.org/2011/prov/wiki/Meetings:Telecon2011.10.27

<pgroth> PROPOSED to accept the minutes of Oct 20 telecon

<dgarijo> +1

+1

<Yogesh> +1

<satya> +1

<SamCoppens> +1

<pgroth> ACCEPTED Minutes of last week

pgroth: 1 action, being updated, to be discussed later
... need scribes

PROV-O

satya: covered new stuff, including new extensions previously discussed

<dgarijo> we've also included examples to all the properties that didn't had them

<scribe> ... new examples: usage of time

UNKNOWN_SPEAKER: no roadblocks to moving to release draft
... need to proof-read end-to-end
... a lot of changes, but content is largely in there now

<dgarijo> +q

dgarijo: content is there, but need to determine approach to modeling

<khalidbelhajjame> Daniel is talking about EntityInRole

dgarijo: How will EntityInRole be handled

<khalidbelhajjame> +q

<khalidbelhajjame> -q

pgroth: Is there a consensus on that in the working group?

<pgroth> in the prov-o working group

khalidbelhajjame: we haven't had a specific discussion on that in the working group yet
... There are a number of issues on mapping the models to one another

<khalidbelhajjame> +q

khalidbelhajjame: discussions continue, there will be modifications. We really need to be sure there are no show stoppers

<dgarijo> -q

<dgarijo> @Khalid: +1

<satya> @Daniel and Khalid +1

khalidbelhajjame: people in the working group working on it,

pgroth: can this be resolved quickly?

<khalidbelhajjame> Paul, the answer to your question IMO is yes

<dgarijo> Both approaches are clear. Both have advantages and disadvantages. We just have to go for one, and stick to it.

satya: basic issue is data model from Luc and Paolo. We are trying to model a language
... it is hard to get a perfect one-to-one mapping

<dgarijo> what imo is not the right approach is have the same discussion every two or 3 months.

satya: Especially with something like OWL
... e.g. what is the difference between annotation and attribution?

<pgroth> don't they have different names?

satya: We need some buy-in to move from ASN to OWL
... what does it mean for provenance? Are we losing something with this mapping to the data model
... There can't be a perfect mappign

<dgarijo> @pgroth: some people argued that summaryOf, for instance, was very domain-specific.

<Luc> satya, are you sayiing interoperability is impossible????

<pgroth> Sub-Topic: Relation between PROV-DM and PROV-O

<pgroth> http://lists.w3.org/Archives/Public/public-prov-wg/2011Oct/0185.html

pgroth: how do we resolve mapping between models?

<pgroth> +q

pgroth: There may be a mismatch. What process should we adopt?

<satya> @Luc: no Luc, the interoperabilty is the aim of the the whole Semantic Web technology stack - hence by using RDF and OWL we are facilitating interoperability

<Luc> i can hear

luc: khalid indicated there was an outline of a solution to address this

<khalidbelhajjame> +q

<dgarijo> +q

<dgarijo> -q

luc: not sure if agreement was reached

<satya> @Luc, so if we are talking about interoperability of PROV-DM and PROV-O it will require us to understand how both the ASN and RDF, OWL relate

khalidbelhajjame: Not everyone responded to proposal

<stain> had to put the baby to sleep

khalidbelhajjame: By adding shortcut properties, we can connect entities to handle this

<Luc> @kahlid, this looks good to me

khalidbelhajjame: we are still discussing this
... topic for next monday's telecon

<dgarijo> +q

<Luc> we lost you , paul

<satya> Paul we lost you

stain: we had this discussion and chose the current approach

(sorry daniel, not stain)

<khalidbelhajjame> @Daniel +1

luc: Need to decide and formalize the result and take that into account
... the current approach isn't good enough

<satya> @Daniel +1

<stain> (what is it we are talking about now, provo at all or something else?)

luc: we could release the draft, but it really has problems, it doesn't match the model

<khalidbelhajjame> Stian, we are talking about Prov-o

luc: we must introduce entities to make it match

<khalidbelhajjame> and EntityInRole

<stain> ah\

<satya> @Luc, I guess what Daniel is saying that we use examples to demonstrate how one or the other approach does not work

<stain> what was that good name that was suggested as an alternative by Jim (I think)

dgarijo: we can try to address that, but we need consensus on how to address it

luc: there are other variants around OPM-O, there should be a way to handle it
... it would be nice to contrast the two approaches with examples

<stain> +1 to try the different approaches with examples

<stain> the same for prov-dm

<satya> @Luc, from my perspective creating named classes for properties does not work in OWL, RDF

luc: can you formulate examples contrasting the approaches?

<dgarijo> currently, all the examples are using the incorrect approach.

<smiles> @Luc, next agenda item is Prov-DM and the two proposals you put on the mailing list (regarding entity, process execution)

Can Paul and Luc join next working group telecon to help resolve approach?

<khalidbelhajjame> +q

satya: there are issues trying to model. You need extra propertities to link the entities
... This will introduce addition blank nodes, they make it hard to transfer between applications

<pgroth> +[IPcaller] is me

<dgarijo> @Satya: they don't have to be blank nodes!

satya: there isn't a good way to transfer the RDF blank nodes

<dgarijo> they have to be unique instances of the properties, just like the unique instances of EntityInRole.

<dgarijo> in fact, the approaches are very similar (in essence)

satya: there are practical reasons why the biomedical community has gone to using roles
... and have a different way of modeling some things

<sandro> @Satya FYI the RDF WG has decided on a standard way to do Skolemization, although it's still just in a Working Draft. not sure that helps much with this problem.

satya: the simple approach runs into problems others in the SW community have already run into

<Lena> on blank nodes: http://axel.deri.ie/publications/mall-etal-2011ISWC.pdf

<Lena> (paper nominated for best paper award as ISWC 2011)

Luc: Maybe that problem doesn't actually hit us, let's look at examples

<stain> you can name any node, obviously.

<dgarijo> they cab be blank nodes, but they don't necessarily have to.

khalidbelhajjame: In our context, we won't have that problem, we have some shortcuts
... that can link the constructs.

<Luc> i hope to be able to join you on Monday

<dgarijo> +q

khalidbelhajjame: we won't have the drawbacks that others have run into

<satya> @Daniel: In OWL, RDF, there is a single URI for properties (both object and datatype) - so a property does not have instances as OWL classes

khalidbelhajjame: we will discuss at monday's telecon

dgarijo: If there are n-ary relationships, there isn't a good solution. The current approach is confusing.

Luc: We will have to resolve this and make a decision.

PROV-DM

<dgarijo> @Satya: but n-ary relationships are a well known ontology pattern too.

<satya> @Daniel, so I am not sure I understand your point

Luc: On monday we will discuss examples of two approaches and converge toward solution.

<satya> @Daniel, can you give examples of ontologies that use n-ary properties modeled as classes?

<pgroth> sorry everyone

Luc: Two proposals were floated on mailing list

<Luc> http://lists.w3.org/Archives/Public/public-prov-wg/2011Oct/0195.html

<stain> the problem is often finding a good name for the n-ary class, we've got Generation, Use and Control, which is not too bad

<stain> they are of course no longer properties

<stain> just n-ary relationship

<satya> @Sandro thanks! I will read the RDF Working Draft

<dgarijo> @Satya: people are complaining because the current approach is hard to follow. And basically, EntityInRole is a "trick" to avoidintroducing the classes. I like the approach because it's simpler

Luc: A number of votes on proposals

<dgarijo> (I mean, the current approach)

Luc: overwhelming support for first proposal

<Luc> PROPOSED: in section 2.1 [1], to define an entity as an identifiable characterized thing.

<dgarijo> @Satya: One example: OPMO :P

<Luc> ACCEPTED: in section 2.1 [1], to define an entity as an identifiable characterized thing.

<Luc> PROPOSED: to rename 'process execution' by 'activity'

Luc: not as much support for second proposal
... rationale - the first proposal introduces entity as an identifiable characterized thing
... we need to bring section 5 to match

<sandro> @satya specifically: http://www.w3.org/TR/2011/WD-rdf11-concepts-20110830/#section-skolemization

Luc: entity expression describes entity

<satya> @Daniel: OPMO has multiple OWL specific modeling issues and there are no application using OPMO for generating RDF in contrast to ontologies listed in NCBO: http://bioportal.bioontology.org/ontologies

<satya> @Sandro thanks!

Luc: there are negative votes for replacing process execution with activity

<dgarijo> @Satya: yes, there are.

<JimMcCusker> +q

Luc: first proposal being accepted and not second would make document confusing

<pgroth> we can't hear you

<Luc> we can't hear you

<satya> @Daniel, also easier does not mean repeating mistakes already known to the community'

JimMcCusker: Plan B is to use some other word

<satya> @Daniel, can you send pointers

JimMcCusker: there are problems with process as well, but it is ok
... perhaps use event?

Luc: that isn't the current issue

<pgroth> +q

<dgarijo> @satya: http://webenemasuno.linkeddata.es/models_en.html, http://www.policygrid.org/ (But according to what Luc is saying, maybe we should leave this for monday)

JimMcCusker: Borrow the term as an alternative

Luc: There is no proposal to remove term activity. We are just trying to align terms.
... We can make it clear that there is no agency involved

JimMcCusker: Activity implies actor

<pgroth> +q

JimMcCusker: invites misinterpretation

<stain> \=1

<Luc> ack

<stain> plus 1

pgroth: Why didn't this come up earlier?
... We can at least simplify this so we are discussing one term

<stain> I think it did not come up earlier because it was just in the preamble of the conceptualisation which was always about "stuff and things in the world" and probably easy to overlook

pgroth: This proposal is trying to simplify use of terms

<Luc> currently, a minimum of 54 occurrences of the word activity in prov-dm

<JimMcCusker> I didn't raise it earlier because "activity" was being used informally.

pgroth: Other terms are a separate issue

JimMcCusker: Ok, if the proposal is to align terms, we should do that

Luc: proposal not perfectly clear, but the intent is alignment

<stain> alignment can happen from both sides

<pgroth> PROPOSED: align terminology of process execution and activity with entity and entity expression

<Luc> PROPOSED: align terminology of process execution expression and entity expression with activity and entity

<Luc> PROPOSED: align terminology of process execution expression and entity expression with activity and entity, respectively

Luc: Jim, is that better?

JimMcCusker: Yes, that is better, I'll raise the issue

<smiles> +1

<JimMcCusker> +1

<zednik> +1

<dgarijo> +1

<stain> +1

satya: 'align' is to use the new terms?
... we are replacing the terms in the document

Luc: Yes, we are using the new terms

<satya> +1 (differentiating between activity execution and activity specification)

<khalidbelhajjame> +1

+1

<SamCoppens> +1

<YolandaGil> +1

<Luc> ACCEPTED: align terminology of process execution expression and entity expression with activity and entity, respectively

<dgarijo> @Satya: but activity specification isn't Recipe (or recipeLink) already?

<dgarijo> I mean, they are already differentiated, right?

<satya> @Daniel: I did not find anything about OPMO and for PRISMA I did not see how OPMO is used for SIOC, MPEG-7 and other ontologies

simon: people are adding a lot to the document
... intro done, examples coming in
... sections allocated, Tim, Ted contributing sections

<Luc> ted?

simon: The rest is almost there, need some examples from some others

<dgarijo> @Satya: in the webn+1 project is used for describing the evolution of the entities (travel guides). SIOC, MPEG-7 and other ontologies are used to describe those entities further.

simon: Paolo has been checking against the conceptual model
... Satya will check against the formal model

<dgarijo> @Satya: http://webenemasuno.linkeddata.es/page/elviajero/resource/Guide/20040117ELPVIALBV_6.TES

<satya> @Daniel: translation :)

simon: examples to be done soon, waiting on Tim,Ted sections, hopefully draft within the week

zednik: Roles examples, same issues with entity and role

<pgroth> you mean entityinrole?

zednik: in the work example, entity/role problems are still confusing

Luc: hopefully we'll have a better solution on Monday

<stain> q\

<stain> q

pgroth: primer - do the examples have problems that are difficult to explain?

<Luc> ... or are difficult to understand in prov-dm?

pgroth: Are you capturing problems explaining things?

Roles/entities are the most difficult, should be resolved on monday

zednik: also accounts

Luc: We haven't really debated much about accounts yet
... We might need to make changes to PROV-DM for accounts, but we haven't addressed

Building an Example Catalogue

Luc: people writing documents, primer, etc. are making examples
... We should build a catalogue of those examples systematically
... We need something wider to evaluate the various representations

<pgroth> +1

Luc: Should we do that? How should we approach it?

<stain> hard to keep in sync

<dgarijo> @Luc: +1 for the catalogue. Are we going to have more than one RDF representation??

stain: examples aren't complex enough to handle everything
... difficult to make them match everything and stay up to date

Luc: concepts will stabilize, then it will be easier

pgroth: We could do it as an adjunct to the primer
... The primer could link to the examples on separate pages, we could restructure
... into a catalogue. Revisit after things are more stable.
... A single wiki page that links to all the different examples.

<stain> +1

pgroth: Link to blog posts, primer examples, etc. at least get a common list

<satya> @Daniel If you are interested, we can have short skype call right after the telcon now (just reviewed web1ontology.owl)

Luc: Something more formal would help with tool development

pgroth: Use wiki page as a directory, then add more structure/formality later

<dgarijo> @satya: ok

<pgroth> yeah

<pgroth> well the turtle file

smiles: How would that work?

Luc: As an RDF file, turtle, that would work.

<pgroth> yeah

Luc: Some comments with context/documentation

satya: How should we link the RDF/turtle files to the wiki/mercurial repository?

<stain> https://dvcs.w3.org/hg/prov/file/90a007a1712d/ontology/examples/ontology-extensions

Luc: Have a top level mercurial area for examples. If Tim has a structure, we can adopt it

<stain> https://dvcs.w3.org/hg/prov/file/90a007a1712d/ontology/components

<stain> but its only provo now of course

<stain> sorry

<stain> son pulled phone

<Lena> (i have to go to another telco, I will continue to help simon in the primer document + RDF examples)

Luc: A top-level project would help organize it. Stain noted there are PROV-O examples already there

pgroth: Let's proceed, sort out details on email

<stain> some of the components ex are out of date or experimental

<stain> but thart can be cleaned

<pgroth> trackbot, end telcon

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.136 (CVS log)
$Date: 2011/10/27 16:02:53 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.136  of Date: 2011/05/12 12:01:43  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Found Scribe: Curt_Tilmes
Found Scribe: Curt
Inferring ScribeNick: Curt
Scribes: Curt_Tilmes, Curt
Default Present: Curt_Tilmes, Yogesh_Simmhan, Sandro, dgarijo, smiles, Satya_Sahoo, SamCoppens, +1.518.633.aaaa, Yolanda, Luc, stain, pgroth, [IPcaller]
Present: Curt_Tilmes Yogesh_Simmhan Sandro dgarijo smiles Satya_Sahoo SamCoppens +1.518.633.aaaa Yolanda Luc stain pgroth [IPcaller]
Regrets: James_Cheney Christine_Runnegar
Agenda: http://www.w3.org/2011/prov/wiki/Meetings:Telecon2011.10.27
Found Date: 27 Oct 2011
Guessing minutes URL: http://www.w3.org/2011/10/27-prov-minutes.html
People with action items: 

[End of scribe.perl diagnostic output]