Provenance Working Group Teleconference -- 10 Nov 2011

<trackbot> Date: 10 November 2011

<pgroth> any volunteers for scribe?

<pgroth> Scribe: Paolo

Admin

<pgroth> http://www.w3.org/2011/prov/meeting/2011-11-03

<pgroth> PROPOSED to accept the minutes of the Nov. 3 telecon

<saty> +1

<smiles> +1

<Curt> 0 (did not attend)

<khalidbelhajjame> +0 (was not in last week)

<tlebo> +1

<GK> Khalid, are you sure ??PP80 is you?

<StephenCresswell> +1

<jcheney> +1

<khalidbelhajjame> @Graham, not sure

<pgroth> ACCEPTED Nov 3. 2011 minutes

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

<khalidbelhajjame> @Graham, I will leave the call and come back and see

Tim's action presumably taken care of

Paul completed his action (42)

Tim's action was 41 -- closed

<pgroth> Reminder F2F2 Poll: http://www.w3.org/2002/09/wbs/46974/f2f2_options/

we are skipping action 40 at this time

TOPIC PROV-PRIMER

Simon: good contribs but still got gaps
... Stephan , Paolo, Yolanda to contribute to a complete draft by this week
... so that the WG can start commenting
... Stephan creating turtle examples
... Simon to complete the intro
... Yolanda to give it a check
... then Paolo to translate turtle -> ASN
... : accounts still missing. That's because it hasn't settled in PROV-O

<khalidbelhajjame> +q

<tlebo> Account will be defined :-)

Khalid: on account. PROV-O will not have explicit account, as named graphs will be used

<tlebo> Account will be part of the ontology :-)

Simon: still, some encoding of them is needed for the examples

Tim: wil use named graphs, but also RDF to express accounts. So it's going to be both

<Zakim> GK, you wanted to note that ORE uses trix graph as a base class for ORE resource map, which also is a named graph

<Luc> ... and also in prov-dm, accounts need to be finalized

Paul: fine, but work is still ongoing in PROV-O re: accounts, which explains why they are not in the primer at this time

<tlebo> ORE - good pointer?

<tlebo> thx!

Simon: distribution of first draft expected by start of next week

<tlebo> A rough example of account modeling: http://dvcs.w3.org/hg/prov/file/3ba83e9ffa92/ontology/components/Account/different-accounts-can-include-the-same-entity.ttl

PAQ

<smiles> @Paolo shall I take over scribing now?

@simon: yes please, much appreciated :-)

<smiles> Scribe: smiles

<Paolo> I will do my other half next time :-)

GK: Updated PAQ, as agreed for FPWD
... Note, not yet fully proof read

pgroth: Big changes are to align the PAQ with the terminology in DM, e.g. entity
... and a decision about the format of headers for retrieving provenance info for a resource
... and added a section to deal with incremental access to large amounts of provenance
... Also compacted things, referring to DM
... Everyone please look at the document

GK: Closed issue tags in document but not tracker
... Comment from Yogesh about not guaranteed to get identifier of entity in provenance data, so added note on this

Luc: Decide in next telecon whether to release FPWD?

pgroth: Yes, would be good to know if there are any show stoppers by next telecon

tlebo: Will predicate hasProvenance be encoded in Prov-O?

GK: Was included for discussion, but yes need to agree with other task forces (namespace, name, inclusion in ontology)

tlebo: Will start developing inclusion of hasProvenance into ontology

satya: What is domain and range?

<tlebo> owl:Thing.

GK: Domain is entity, range to be decided (account?)

<tlebo> ProvenanceContainer?

satya: What provenance is may change across applications, need to assert about account or container itself

GK: Yes
... Account or container is itself an entity

<satya> @GK +1 for that point

<dgarijo> @GK that sound good to me too

Update on PROV-O

satya: Fleshed out details on how to add qualifier info to predicates, modelled under class QualifiedInvolvement
... Outstanding issues: need good name for QI to entity link

<tlebo> (we have been running with prov:entity, but prov:entityInQualification was suggested and sounds reasonable)

<Paolo> apologies for checking out now --

satya: inference rules to apply to non-binary properties with new classes
... need clarifications on DM: can roles be associated with both entities and process executions?
... at the moment only one or the other
... Moving forward, all terms except "entity in role" modelled, so working towards FPWD

<tlebo> The proposal is at http://www.w3.org/2011/prov/wiki/Qualifed_Involvements_in_PROV-O

satya: n-ary properties modelled as classes, can say 8 as denominator in division process
... but cannot model role of process execution

<zednik> the process execution has a role (part or function) in itself?

Luc: Please send an email explaining problem with example

<GK> That ORE reference I mentioned for mentioning graphs in an ontology: http://www.openarchives.org/ore/1.0/vocabulary.html#rem

satya: sure

<tlebo> Thanks!

PROV-XML

pgroth: In charter, have notion of natural XML serialisation of the DM
... due at 18 months, but can start thinking about now
... want to know who is interested in starting to produce this serialisation

<pgroth> +q

jcheney: Some people previously said that RDF can be expressed in XML, but sounds like in charter going straight from DM to XML
... would be interested in being involved in some way

pgroth: Yes, in charter, straight from DM to XML, RDF/XML is not pretty XML

<GK> Presumable, want something that plays well with XML tooling, which RDF/XML does not.

jcheney: We should agree that this is indeed what is intended

Luc: Interested in this, have had questions from users on OPM XML and interested in Prov XML schema, and they may be interested in contributing
... Has very early attempt at XML schema

pgroth: 2 people interested, maybe need to bring in other experts from outside

<zednik> xml - I may be able to help, but will not be able to lead

<GK> I might be interested in JSON :)

<khalidbelhajjame> Me too Graham

Luc: questionnaire circulated showed interest in many serialisations, so some may be able to help with XML

<Curt> +JSON

zednik: Yes, users interested in XML
... close to that of RDF

<GK> (Even Zakim is interested, apparently :) )

Luc: go back to those people?
... First go back to those people for feedback

<tlebo> @gk, could you write something at http://www.w3.org/2011/prov/wiki/Using_named_graphs_to_model_Accounts#Graham.27s_OBE_note ?

zednik: will do so

Luc: may not be able to list on Wiki or email for privacy

zednik: some users agreed to have feedback shared, can put document up on protected W3C site

<GK> @tlebo, sure

pgroth: Saw JSON interest on IRC, are people interest?

<pgroth> Interest in JSON note?

<satya> +1

<zednik> +1

<khalidbelhajjame> +1

<sandro> +1

<jcheney> +0.5 (what would we say exactly?)

<GK> +1 ... but not in a rush to do it

<Curt> We use JSON internally, but I think RDF makes a better standard for interchange.

<Curt> +1

<dgarijo> +0

PROV-Semantics

pgroth: Deliverable on semantics in charter, but up to us to decide what is usable and interesting for standard

<jcheney> Two possibilities (not mutually exclusive):

<jcheney> 1. Developing a mathematical model of the "things", "entities", "processes", "events" and other relationships as in the PROV-DM, and explaining the PROV-DM statements in terms of this model. (current strawman)

jcheney: Current strawman generated some discussion, but died down, also needs updating to current DM
... Luc said was helpful, can provide some justifications for inferences

<jcheney> 2. Defining the mapping from PROV-DM to PROV-O (and maybe "PROV-XML") formally, e.g. using a datalog or ML-like notation.

jcheney: We might not just want to specify data model and serialisations separately, but also formally how we map from DM to those representations, what it means to be a correct translation
... Don't want to have multiple translators between each pair of serialisations, want to translate to Prov-DM and back

<Zakim> GK, you wanted to say that I think there's some confusion around DM, ASN and semantics

GK: Concerned that there is a confusion between the DM and the RDF representation (as James said, but focus more on concepts)
... concerned about pushing RDF concepts into DM without RDF semantics, better for DM to be above the RDF structure

<satya> @GK +1 for not conflating DM and RDF semantics'

GK: formal semantics, independent from OWL, for DM could help with this
... then may be possible to prove that RDF semantics corresponds to abstract DM

<tlebo> +1 for adding a DM semantics. Some of the inferences in the DM writeup are difficult to follow from its narrative.

satya: Not clear how mapping is related to formal semantics, why not just translation

<GK> @satya - isn't this like prrof-theoretic and model-theoretic laters?

Luc: James' suggestion 2 is good from interoperability point of view, regardless of whether part of formal semantics activity

<satya> @GK, I guess but not sure in context of DM and its semantics

jcheney: We already talk about how to translate ASN to Prov-O in Prov-O document, so thought useful to have more mathematically precise defn of that in formal semantics

<Zakim> tlebo, you wanted to ask about mechanics of a concrete language for DM, and it's mapping to XML and RDF and JSON.

<satya> @James - I think we need it

jcheney: if we have one deliverable of formalisation, then a formal mapping to serialisation should go there

tlebo: How does mechanics of formal semantics work? How different to, more precise than the serialisations?

<satya> @James - In addition, as WG we have the responsibility for defining the mappings between the different representations (DM, PROV-O, XML, JSON)

jcheney: First thought of what goes in formal semantics is like RDF semantics, e.g. what you can write in the language

<tlebo> what do you mean by "scope" :-)

<tlebo> naming or account partitioning

jcheney: If we have semantics abstracts from what you have to write down, then can express self-consistency of scoping rules etc..

<Zakim> GK, you wanted to respond to luc - I think there's a difference between interop and provable equivalence of representations/transforms

GK: In SW area, model theoretic semantics maps OWL/RDF expressions to objects in domain of discourse (set theory)
... With regards to interoperability, difference between demonstrating interoperability and formally proving equivalence
... Pat Hayes formal semantics of RDF is a useful intro to model theoretic semantics

satya: Important to define mappings from DM to serialisations, but how necessary to define semantics of DM/ASN itself? Is outcome that we are defining a new language, ASN?

<Luc> we would give the semantics of DM not ASN!

pgroth: Some agreement for a need for formal semantics of DM (suggestion 1 by James)

<Luc> we would give the semantics of DM not ASN!

<Paolo> @satya: the semantics is of the model not the language

jcheney: To move forward, first need to catch up with DM and compare with strawman
... regardless of whether mapping is formal semantics or not, still clear it is useful and focus on first

<satya> @Paolo: I will reserve my comments (till we have more details of the formal semantics of DM means)

<GK> James mentioned a datalog approach: I think that could be used to build in formal semantics from FoL - for which there exists a model theory.

<satya> @GK, ok that makes sense

jcheney: Also happy for anyone interested to be involved, starting with mapping from Prov-DM to Prov-O

<GK> There was a proposal by R V Guha and (I think) Pat Hayes, many years ago, to do something sikilar for RDF.

<satya> @James - I can help you with that

Prov-DM

<pgroth> Proposed: Use a single notion of attribute-value pairs to characterize entities, activities, use and generation. As a result, drop the notion of qualifier and its associated production."

<Paolo> @satya: set-theoretical interpretation is usually what works with data models

<pgroth> http://lists.w3.org/Archives/Public/public-prov-wg/2011Nov/0122.html

<GK> +1 (but have separate concern about the phrasing using "characterozation")

pgroth: any objections?

<dgarijo> +1

<Paolo> +1

tlebo: what was the intent of the distinction?

<jcheney> @satya, @paolo: The strawman is an attempt to map PROV-DM in terms of sets/functions.

<Paolo> good, thanks

Luc: Attributes were in context of entities, fixed in characterisation interval; relations did not have durations

<jcheney> @GK: Yes, datalog is interpretable in terms of FO model theory; however, dealing with things that change over time seem hard to model this way. Still, datalog good as a lightweight formalism.

Luc: but distinction did not bring much, so better to merge

<GK> @jcheney if functions themselves are sets of pairs, that maybe starts to look like a model theory?

<pgroth> Accepted: Use a single notion of attribute-value pairs to characterize entities, activities, use and generation. As a result, drop the notion of qualifier and its associated production.

<pgroth> http://lists.w3.org/Archives/Public/public-prov-wg/2011Nov/0087.html

<tlebo> Tim's notes on Luc's response: attriubtes on entities (duration, characterization, etc) same for PEs. but for Relations (didn't have durations). ATTRIBUTE-values were for Entity+PEs, NAME-values were on Relations.

<jcheney> @GK: Correct, using functions doesn't take us out of set theory/model theory semantics.

<satya> @Paul: I also need additional clarification

pgroth: Fairly well accepted, except for Simon's objection

<satya> @Paul: I did not have time to respond to this issue

Luc: actually very few voted either way
... we haven't got enough support yet to resolve here, need to understand what Simon is saying

<GK> I was unclear about dependedUpon/eventuallyDerivedFrom distinction.

<tlebo> I'm confused by the use of multiple proposals; will try to read and comment on email.

<tlebo> (but I did get the impression that much of those predicates were redundant)

<GK> I think the transitivity issue is a different one

Luc: we need a notion of transitive derivation, good examples of non-transitive when linked to activities, but unclear on wasEventuallyDerivedFrom

<Paolo> general proposal: in addition to recording objections on the list (which may have veto effect), keep an exact count of the people who vote on the list -- the support to a proposal

<GK> simplification is good!

pgroth: Goal is to simplify

<pgroth> trackbot, end telecon

- DRAFT -

Provenance Working Group Teleconference

10 Nov 2011

Attendees

Contents

Admin

PAQ

Update on PROV-O

PROV-XML

PROV-Semantics

Prov-DM

Summary of Action Items

Scribe.perl diagnostic output