Provenance Working Group Teleconference

15 Sep 2011


See also: IRC log


Paul Groth


<trackbot> Date: 15 September 2011

<pgroth> anybody up for scribing?

<pgroth> can I get a scribe?

<pgroth> can I get a scribe?

I can scribe

if people not on the queue remember to say their name ;)

<pgroth> thanks stain

<pgroth> Scribe: stain


pgroth: Finish within 1h due to RDF provenance telcon afterwards

<pgroth> http://www.w3.org/2011/prov/meeting/2011-09-08

<pgroth> PROPOSED to accept the minutes of Sep 08 telecon

<smiles> +1


<Curt> +1

<Paolo> +1

<kai> +1

<tlebo> +1

<khalidbelhajjame> +1

<pgroth> Accepted Minutes of last weeks telecon

<pgroth> http://www.w3.org/2011/prov/track/actions/open

pgroth: Action items to review
... no actions

<khalidbelhajjame> sorry Paolo

<pgroth> http://www.w3.org/2011/prov/wiki/Scribes

<Paolo> not sure who I am :-)

pgroth: Need more scribes, please sign up so we don't have to assign

<pgroth> http://www.w3.org/2011/prov/wiki/ProvenanceRDFNamedGraph

ITEM Named graphs requirements

pgroth: several people ave signed up for this telcon. Any comments on the requirements?

<Luc> who will joing the call?

<pgroth> +1

<Luc> +1

<kai> +1

<Paolo> +1

pgroth: Say +1 if you are attending the call

<Luc> satya?

<MacTed> +1

<satya> Hi Luc, I am here

pgroth: the call is immediately following this call

<Luc> will you join rdf call?

<pgroth> http://lists.w3.org/Archives/Public/public-prov-wg/2011Sep/0073.html

Thursday 15 Sep, 1215pm US Eastern time for 45-60 minutes 18:15 Paris/Berlin/A'dam; 117:15 London)

<pgroth> Call agenda: http://www.w3.org/2011/rdf-wg/wiki/Meetings:Telecon2011.09.15

<satya> Can we give examples from previous work?

Luc: Concrete examples of where we need named graphs. We don't have concrete examples at this point in time. Wanted to ask members like Satya and members working with (?)

when would we have a serialisation to RDF where we can discuss the need for named graphs?

^^ Luc

satya: we can create a usecase for named graphs directly
... we have previous examples from biomedical domains, requiring named graphs to refer to a set of provenance assertions

both examples can be given

Luc: as a working group we need to decide that indeed this is the way we want to do things. We may need an internal discussion before telling the RDF WG
... to avoid misleading them

satya: could we have an example on the provenance ontology wiki page?

Luc: perhaps that, yes

satya: will create that and put it up

Luc: do this as agenda item for next week?

<scribe> ACTION: Satya to Do named graph example on provenance ontology page [recorded in http://www.w3.org/2011/09/15-prov-minutes.html#action01]

<trackbot> Created ACTION-39 - Do named graph example on provenance ontology page [on Satya Sahoo - due 2011-09-22].

kai: Dublin core metadata provenance group, comments on collective requirements.
... Ability to retrieve the provenance of an RDF resource is required.
... main thing about named graph is taht we can retrieve provenance about RDF statements
... this can be misinterpreted as te provenance of the resource (given by the URI) which we can do directly with RDF

<Luc> @kai, are your requirements explicit in the requirement page?

<tlebo> Will we be adding the named graphs examples to http://www.w3.org/2011/prov/wiki/ProvenanceRDFNamedGraph ?

satya: Responding to Kai - on ability to refer to aprts of provenance
... distinction to bring up, named graphs and reifications allow you to make assertion on statement level
... which would let you refer to provenance of RDF subject, predicate and resource level
... named graph would only give you the granularity of statements

<tlebo> +1, didn't quite follow Satya's distinction.

kai: not sure when that granularity would be helpful

satya: would explain tis on the wikipage

<tlebo> difference between an RDF statement and its S, P, and O.

satya: provenance context entity, google that - example scenario
... need to explain the point of why..

?: about kai's requirement, could you put that there?

Zakim: who is speaking?

<Zakim> kai, you wanted to ask for an example

<satya> @Mac: I don't think there is a difference

MacTed: what is the difference, if the resource is a building, brick, etc.. granularity requirement for an entity should be the same

pgroth: collection of smaller things

kai: you want to describe provenance of something, at least you have a good possiblity to identify a set of RDF statements with named graphs. Reification, yes, but you can't directly talk about a set of statements because you can't identify them. But I don't see this to have antying to do with granuliaryt

MacTed: should reword requirements 2 to "Ability to retrieve the provenance of a set of triples"

pgroth: Kai and Satya has different requirements - we might not understand Satya's reqs which he will clarify
... we'll discuss this afterwards

TOPIC Name for the standdar

pgroth: Moving towards PROV - Luc can explain

<tlebo> what about named graphs needs to be handled as something more than files in a directory?

Luc: Last week's telcon there was strong support for the name "Prov"
... this was put out on email last Friday, but not received much feedback except from GK whi did not oppose it

RESOLUTION: Name was decided as Prov / PROV (casing not decided)

<sandro> It's just a name; I wouldn't all-caps it.

Agree - we said last week that it was not a acronym

First working draft

pgroth: Time table for this. GK is not on the phone.


<pgroth> Yogesh

Yogesh: nothing to add

Luc: In last weeks call, we are still aiming to release by end of month - to do this we need a resolution by the group that we are willing to release the document as working drafts
... would like to have the documents approved on the 29th in 2 weeks time
... to do so we will finish the model document this week, ontology document following soon. Wanted to know if PAQ document would follow same time table

pgroth: hangup on PAQ document is dependent on conceptual model
... GK has emailed that we need to have those terms clearly defined in conceptual model
... don't know the details. Likely we can follow the same timeline, but a week later for PAQ
... Any other comments?

Formal model document

satya: discussion on role
... call on Monday, discussing how to model roles and how to interpret them in our model

pgroth: that's te next discussion point

<scribe> agenda: http://www.w3.org/2011/rdf-wg/wiki/Meetings:Telecon2011.09.15

pgroth: working on extensibility of prov ontology

^^ Satya

satya: how different domains can extend ontology, doing concrete examples
... to see if we can make Taverna example as an other usecase to deminstrate extension with new classes and properties for scientific workflows

<Paolo> +1 for using the Taverna example

Luc: Is this document to become a normative document, is it then appropriate to have an example for specific technology like Taverna, or a more neutral example
... Perhaps don't specify this as part of the specs

<tlebo> can we accumulate tool-specific concrete examples on the wiki?

satya: take out Taverna specific details, but follow the scenario in a general way

<dgarijo> I agree with Satya. It is just a Taverna workflow, but could be any scientific workflow system

stain: Would not include specific Taverna-details, but do a general simplified example for scientific workflows - good because one can also show a diagram of the abstract workflow

pgroth: so build another example from the Taverna example, but do a general one?

<dgarijo> @stain: +1

satya: perhaps just a diagram on how Stian could hae extended the ontology, and some explanation

ilkay: Could also try to validate this from the Kepler point of view

satya: that would help a lot - you could work with Stian

Ilkay: Will contact Stian

Roles and times - how they can be associated with Used and Generated

satya: could Luc bring up the initial issue?

<pgroth> ace Luc

Luc: Conceptual model defines a type of relationship, Process execution Used an Entity, or an Entity was Generated by an PE
... and there are some properties to those relations, like te notion of "role" which we just call a qualifier in the model, describing the type of interaction

not just binary relation, an n-ary relation

Luc: Back some years ago in an early OPM serialisation, these n-ary relations was exposed as resources

there were some comments that it was not a very natural RDFisation

Luc: OPMV used RDF properties to express those relations
... Which is fine if you don't talk about roles and times together with Use/Generation
... But what if you want to do this, how would you do this in RDF

satya: what we discussed was to specically have a class Role, we have been discussing how to model this
... we can use the approach of where a Role is a special type of Entity
... in the example of Khalid - Khalid as a person, say as a researcher at Univ of Manchester
... but Khalid at a restaurant is te role of Customer
... or play football, where he assumes the role of a GoalKeeper
... the specialisation that Luc described in a model perspective
... we are then pushing the specialisation from the property to the entity itself
... Khalid can assume these different roles
... we can relate entities to these roles - and on the role we can assert things like time, etc.

Paolo: We had a brief discussion with Satya and the rest of the group

<GK1> So what is he at a restaurant talking research with colleagues?

<tlebo> I'm not sure we need to relate the Used entity with a distinct Role - Why not put the role directly on the Used Entity?

Paolo: not a relationship, but a persona, an Entity assumes this for the duration of this action

<khalidbelhajjame> +q

Paolo: temporarily assocated to entities by way of specialisation, interesting, but departure from model

<dgarijo> @tlebo: you could do that by specializing used, but the role is a trick to model the n-ary relationships

<satya> @GK: Can you please clarify

<tlebo> BTW, the notes from the OWL telecon are at http://www.w3.org/2011/prov/wiki/PIL_OWL_Ontology_Meeting_2011-09-12

khalidbelhajjame: agree with Paolo.
... relationships we are describing are with relation to attributes
... if we want to map this, we need to define the relationship in the contextual model as first class citizens

<tlebo> re "role is a trick to model the n-ary relationships" - that is fine and a Good Thing. But let's put the n-ary directly as the Entity that is used by the PE.

khalidbelhajjame: Luc said someone stated this as a bad idea.. but..

<GK1> @Satya - I was thinking that it has been said that there can only be one role used - so if it's applied to the "person", which applies?

khalidbelhajjame: if we can't define the relationships as classes in OWL

<Luc> wasGeneratedBy(e1,pe1,qualifier(port="p1", order=1),t1)

<satya> @GK - no I meant multiple roles can be used

<dgarijo> @tlebo: it is modeled that way already

<GK1> (@Satya - being late joining, I may be missing the point.)

@GK1 no, it should be possible to use it in different roles in same PE?

Luc: Value of the entity at a given port - ordering

<GK1> @Satya, @Stian: Ah, OK

Luc: one example we want to support in the model

@GK1 Multiple generation roles for same entity is more interesting :)

Luc: Role might have been misunderstood - not like in role-based access controlled
... It is given information about the actual usage in the system

<dgarijo> @stain, I think that with this approach it is covered too

Luc: not sure about satya's notion of Role as subclass of Entity

@dgarijo I believe so too

<satya> @Luc: yes, we can model the qualifiers using roles as we discussed

<dgarijo> @stain: they would be 2 roles used by the pe and assumed by the same entity

smiles: about expressibility (???) - has relationships of roles and time information

(could someone fill in first bit of smiles argument?)

<khalidbelhajjame> +q

Paolo: supportive of example Luc gave, good on general req to codify this relationship which won't go away. smiles idea is sensible - two-layer approach where you can express this or not
... interesting as Satya described it - for the duration of an activity, an entity assumes a persona/role - but I'm afraid..(?) this example. could Satya explain?

<smiles> my argument was to have 2 ontology representations: one is intuitive, maybe relies on reasoning but lacks expressivity; the other allows expression of time on edges etc. but relies on "used" etc being classes

<tlebo> @stain - smiles' two layers?

<smiles> (a la OPMV and OPMO)

khalidbelhajjame: RDF simon of having two versions - like the notion of roles. If we want to do this properly will not appear in the simplified version, it qualifies the relationship

<tlebo> I don't think it's about simple vs. complex, it's about whether the extra context (role, time) is asserted on the used Entity or not.

khalidbelhajjame: would it be sensible to have the simplified version in the ontology

@tlebo that makes sense

satya: not two versions of ontology, Role should be part of ontology
... question is what the information we are trying to represent
... statements on the entity or on the process execution

<tlebo> satya: two distinct things: (I missed the intro)

<tlebo> qualifier on the relationship vs. qualifier on the entity.

satya: say entity on port 1, ordering 1 - are these properties on the entity itself - qualifier on the entity, then modelling roles as entity allows us to say this
... that entity was the first package on a port

<Luc> why is it a qualification of the entity? it's not an attribute of the entity?

@Luc agree - and an entity can be used for multiple roles wit different properties

like a hammer used both for hammering nails and pulling them out

<Luc> @stain, indeed

(but you could say those are two views of the hammer?)

<tlebo> Does this work? :my_pe prov:used [ a prov:Entity; prov:actually :Khalid; a prov:Role, a restaurant:Customer, time:begin :t1, time:end :t2 ] ?

Paolo: (?) complex bit you need to make explicit. that data was produced.. (?)

@tlebo that is satya's proposal, yes

@tlebo kind of like ORE proxies

satya: (..) customer left the restaurant at this point in time, etc.

Paolo: we don't have this in the abstract model

satya: possiblt need to bring this up to the WG

<tlebo> @stain, thanks, I agree with this approach. Before Monday's telecon with Luc, I conceived of Role and the used Entity as distinct (but I don't like that difference without a purpose).

Paolo: like the idea of qualifying entities, bu tneed to bring this into the language and discuss this

pgroth: no final agreement, but conversation! Need to move on on the agenda

TOPIC How can we identify attributes of an entity

Luc: Identify an entity and attribute (key-value pairs)
... these describe something constant int he world during the duration of the entity's existence

<dgarijo> if anyone is interested to particiate, we have our ontology telecon on Mondays :)

Luc: Need to know which attributes have been "stamped" on the entity to characterise it
... Don't know how to find these attributes with the OWL mapping
... Some examples were discussed, Stian had one proposal, but don't know if this has been incorporated

<khalidbelhajjame> +q

<GK> @Stian, that sounds like reading too much into anonimiy of a node

<tlebo> The entity need NOT be a bnode/anonymous. It can be named with a URI (the bnodes in examples are a shorthand).

<GK> ... you can assign a new URI a an anlymous node without changing the meaning

<tlebo> *used Entity

Stian suggested :entity :charactizedBy [ :location "Manchester", :colour :red ]

<Luc> we could use named graphs to "wrap" the attributes

khalidbelhajjame: to introduce Properties or Attributes into the formal model - or characterized-by, descibed-by
... then it can be instances of this - distinguish characterized attributes and other supplemental

<tlebo> How would :entity :charactizedBy [ :location "Manchester", :colour :red ] handle :entity prov:?? <http://tw.rpi.edu/instances/TimLebo> . (MY _actual_ URI,not a description of me)

khalidbelhajjame: what is the scenario given - most of the time attributes on the entity will be part of characterizing it

<GK> @Stian :entity :charactizedBy [ :location "Manchester", :colour :red ] ; [ :location "London" ; :color :blue ] . is also valid?

@GK - no, it has granularity 1 so that those nodes would be merged

khalidbelhajjame: not quite clear yet..

Luc: Might have a series of properties your thing has. Like a colour

<GK> @stian quite - just clarifying.

Luc: (car colour example)

@GK it is an important point

<tlebo> :entity :charactizedBy [ owl:sameAs <http://tw.rpi.edu/instances/TimLebo> ] . # would fit, but is a bit indirect.

Luc: It is an active assertion by the asserter to say that some attributes were constant.
... The asserter might not care about colour, but talk about registration of the car
... Although the colour is recorded, it might not be part oft he characterisation made by the asserter
... We want to distinguish what the asserter says characterizes an entity or other props

khalidbelhajjame: so someone else added the colour attribute?

Luc: right

pgroth: how to write this down in OWL.. given the time

<tlebo> Luc's point about distinguishing between assertions of provenance maker and OTHER assertions about the same thing - this is handled by placing those attributess on the used :Entity, no?

pgroth: we can keep discussing this on mailing list and move on to conceptual model agenda item

Conceptual Model

<GK> Update recent?

Paolo: moving forward with Luc
... on track for internal release tomorrow

<Luc> we should go for Monday release, realistically

Paolo: few things in flux, a section on providing a high-level overview of model
... working on that
... adding a more precise description on what we mean by collections and relationships to support collection membership

Luc: spent some time thinking about entities, following issues/emails by GK

<GK> Good, I look forward to seeing the update.

Luc: we came to a resolution here, a reasonable way to talk about entities
... Using them in the document

<GK> "here" is earlier this telecon?

(sorry I am not sure)

Paolo: discussion on Account - coming along
... shift in view from Roles and Attributes - perhaps most of the things you talk about can be qualified by attributes (key/values)
... some extension point
... one way to extend the model is to add attribute value/pairs to a profile for instance
... define how those are used
... one consequence is the discussion on wasGeneratedBy
... also on Account
... can be nested inside each other - scoping rules
... getting complex.. giving ourselves a few more days

pgroth: any questions

<khalidbelhajjame> +q

<Luc> "here" was "at our meeting Paolo and I"

<GK> @luc thanks

khalidbelhajjame: in two weeks time would like to have.. (? )
... how would this work - we raise issues towards the doc in one week and other week..? We only have two weeks!
... should plan how to manage issues
... to make it for the deadline

Luc: to raise issues with the tracker
... realistically we will not address them all by end of Monday or the 29th
... there will still be work to be done
... want to have it in a state where we can say it is our first public working draft with clearly identified/marked issues

satya: can we also have a (?)
... if Luc/Paolo meets to have a telcon

Luc: meeting Paolo in London next week
... rest by email
... can schedule a telecon if that is wanted

satya: or just a skype call so we can listen in

Paolo: we don't have a regular call, but can set one up
... or join your ontology call on Mondays

<satya> great thanks!

pgroth: ok, need to end now for next telcon! (RDF WG)
... see you all next week

<dgarijo> goodbye!

pgroth: will you do the magic bit of the wiki?

<sandro> quick break before rdf/prov telecon!

<pgroth> trackbot, end telcon

Summary of Action Items

[NEW] ACTION: Satya to Do named graph example on provenance ontology page [recorded in http://www.w3.org/2011/09/15-prov-minutes.html#action01]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.136 (CVS log)
$Date: 2011/09/15 16:00:11 $

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: stain
Inferring ScribeNick: stain

WARNING: No "Present: ... " found!
Possibly Present: Curt Curt_Tilmes Duncan GK GK1 Luc MacTed OpenLink_Software P0 P14 P15 P38 P4 P45 P46 P49 P5 P61 P65 Paolo Satya_Sahoo StephenCresswell Vinh Yogesh Yolanda aaaa aabb aacc aadd aaee dcorsar dgarijo ilkay jcheney joined kai khalidbelhajjame pgroth prov rgolden sandro satya smiles stain tlebo trackbot zednik
You can indicate people for the Present list like this:
        <dbooth> Present: dbooth jonathan mary
        <dbooth> Present+ amy

Agenda: http://www.w3.org/2011/prov/wiki/Meetings:Telecon2011.09.15
Found Date: 15 Sep 2011
Guessing minutes URL: http://www.w3.org/2011/09/15-prov-minutes.html
People with action items: satya

WARNING: Input appears to use implicit continuation lines.
You may need the "-implicitContinuations" option.

[End of scribe.perl diagnostic output]