RDF Application Profiles Group Meeting, Jan. 20, 2014

PRESENT:

Adrian Pohl, Antoine Isaac, bernard.vatant, cirq-kai, e_droege, Gordon Dunsire, Karen Coyle, marianamalta, Thomas Baker

ACTION ITEMS:

  • Stephanie: Set up Doodle poll for next meeting in about one month
  • All: Add use cases to wiki, to be discussed at next meeting

Introduction by Kai: need for a formalized AP which can be validated against; to serve as documentation (internal and external), and validation rules.

DISCUSSION SUMMARY:

Tom: Nervous about having essential information "out of band"

Karen: Need to start with a definition of Application Profile and determine requirements to guide our work.

Kai: When mixing and matching existing vocabularies, when you de-reference you get the standard definition. There's no way for you to provide comments. Also, you may have to access a large number of vocabularies to get the full set for a single application.

Karen: (points to DC guidelines document http://dublincore.org/documents/profile-guidelines/)

Adrian: Points to work being done on Lobid.

Current profiles are in PDF documents, but even so do not follow a particular methodology

Gordon: Would like an RDF representation of DCAP with DSP. Want to return a Concise Bounded Description together with a validation profile that would validate within a domain. Something that would express to others information about my data.

Tom: There can be more than one view per application; AP should focus on view. Also, a single format should be able to provide documentation for humans as well as machine-actionable constraints. Format can help designers write down the constraints they enforce. Provider documentation was not covered at W3C validation workshop.

Kai: You need to be able to define what is a "complete" statement or record.

Mariana: People were frustrated with DSP because it wasn't actionable; there was no way to use it for validation.

SKYPE CHAT

Stefanie Ruehle: sruehle added Adrian Pohl, Antoine Isaac, bernard.vatant, cirq-kai, e_droege, Gordon Dunsire, Karen Coyle, marianamalta, Thomas Baker to this conversation

Stefanie Ruehle: Call started

Karen Coyle: could someone post the link to the wiki page? thanks

Kai Eckert: http://wiki.dublincore.org/index.php/RDF-Application-Profiles

Stefanie Ruehle: sruehle added goldfarb_onb to this conversation

Kai Eckert: http://wiki.dublincore.org/index.php/RDF-Application-Profiles

Thomas Baker: I have a comment

Karen Coyle: tom: nervous when out of band information changes semantics - when someone doesn't have your out-of-band information

Thomas Baker: karen: we have jumped forward - we need to have a working definition of AP and talk about requirements - to set a box around our work

Thomas Baker: kai: mix-and-match existing vocabs. Common concern: if you access terms, get standard definition. No way to provide comments. No way to get the full set without accessing five vocabularies.

Karen Coyle: http://dublincore.org/documents/profile-guidelines/

Antoine Isaac: q+

Bernard Vatant: q+

Adrian Pohl: I'd like to comment.

Gordon Dunsire: http://wiki.dublincore.org/index.php/DCAM_Revision ???

Mariana Curado Malta: I am working on that

Antoine Isaac: Adrian, I would even disagree re. creating a PDF. I don't feel there are really standard gudelines on how to do it...

Mariana Curado Malta: that is my Phd work...

Antoine Isaac: (or at least they are not applied)

Adrian Pohl: yes, I was talking about representing an application profile in RDF.

Mariana Curado Malta: http://dcevents.dublincore.org/IntConf/dc-2013/paper/view/178/81

Gordon Dunsire: We discussed a lot of this stuff wrt DCAM Revision (see link above)

Stefanie Ruehle: DC did some work on a syntax for wikis, see: http://dublincore.org/documents/dsp-wiki-syntax/

Gordon Dunsire: Totally agree - what I would like is an RDF representation, with guidance, of a DC Application Profile Description Set Profile ...

Gordon Dunsire: I want to know how much I can do within an RDF data environment, and how much I have to use other representations.

Antoine Isaac: muted

Gordon Dunsire: I want to return a Concise Bounded Description mashed with a "record validation profile" to dereference an element in my maintenance domain.

Karen Coyle: validation would be built on AP? is AP itself a validation method, or is it a constraint definition?

Mariana Curado Malta: I see it as a constraints definition

Thomas Baker: antoine: with stefanie, working on collecting mappings to Europeana Data Model. A lot of these are in effect APs, but we never defined what we need. We need validation that can work for RDF data.

Thomas Baker: ...there could be several validation strategies. Three versions of XML schema because different constraints at different points of process.

Thomas Baker: kai: for content negotiation on web - place to create comments, documentation - create right out of technical definition.

Thomas Baker: Tom: I would like a way to validate a graph against a "view" that defines what constitutes a "complete" graph and that defines a set of constraints to which a graph should (or must) conform. We would need a language for writing down that set of constraints - something that is easy for data designers to understand and author (as Karen says: "define and document"). That language for constraints would ideally be translatable automatically into a language that could be used for validation (conformance to constraints), such as SPARQL -- a technology for making the validation itself.

Karen Coyle: tom: there may be multiple views; what is a complete view, and what is the conformance of a graph to such a view?

Karen Coyle: tom: the documentation for users should ideally be translatable to an actionable technology

Karen Coyle: kai: providing a basic, local, subjective view of their data

Gordon Dunsire: I just want to describe, in RDF, the context of RDF vocabularies which I publish. I design those vocabularies for a specific (local) application/environment, but I publish them so that 1) they can be re-used globally, and 2) to allow datasets produced in my environment can interoperate with global data.

Karen Coyle: kai: "complete" -- what is a complete statement

Karen Coyle: tom: help designers of data write down what constraints they enforce; w3c workshop was mainly about validation technology, but not about provider documentation.

Gordon Dunsire: So I want to give all other applications the documentation I have for my local applications - for example, the rules for property values, the expected repeatability or mandatory status of elements, etc.

Mariana Curado Malta: TOM: DCAP has the DSP, right? But DSP os not practical...

Mariana Curado Malta: *is

Mariana Curado Malta: "not interpreted by a machine"

Karen Coyle: mariana - there is an xsd for DSP; I'll try to find the link

Adrian Pohl: http://api.lobid.org/

Adrian Pohl: http://edoc.hu-berlin.de/series/dini-schriften/2013-14/PDF/14.pdf

Antoine Isaac: Oh, yes, I agree Adrian :-)

Antoine Isaac: (it's just that everyone ends up doing their own 'style' of doc

Thomas Baker: Mariana: DSP exists, but ideally we'd have a way to make it easy for people to author APs using DSP (something like the http://dublincore.org/documents/2008/10/06/dsp-wiki-syntax/ wiki syntax). Then we'd need a way to turn that DSP representation into a set of constraints that could be used for validation (e.g., SPARQL queries).

Mariana Curado Malta: Karen: but xsd is not a "going back" in technology? (sorry if I am saying mistakes...)

Gordon Dunsire: I want my APs to be machine-actionable.

Mariana Curado Malta: Right Tom. I understand that.

Adrian Pohl: https://wiki1.hbz-nrw.de/x/aIaf

Antoine Isaac: +1 for putting Adrian's proposal for discussion in the group!

Thomas Baker: Mariana: also, the DSP is based on DCAM, which gets in the way a bit because people do not know it (and it may not be

Antoine Isaac: Gordon++ Every project contributing data to Europeana make their AP, roughly.

Antoine Isaac: q+

Karen Coyle: dsp xsd no long on DC site. put it here: http://kcoyle.net/temp/dcmi-dsp.xsd

Karen Coyle: yes, xsd is "yesterday", but this is something one can play with today

Mariana Curado Malta: Antoine: the guy from Madrid was the leader of the VMAP building process (VARIAZIONI Metdata Application Profile)

Mariana Curado Malta: Antoine: he is Carlos A. Iglesias

Mariana Curado Malta: Antoine: if you are interested in his contact I can send it to you

Karen Coyle: do we have a mailing list?

Antoine Isaac: Thanks, Mariana! That's ok, I just to see what they've done.

Adrian Pohl: +1 for 1 call per month

Antoine Isaac: +1

doron goldfarb: +1

Thomas Baker: Karen, see http://dublincore.org/moinmoin-wiki-archive/architecturewiki/attachments/dcmi-dsp.xsd

Bernard Vatant: ok

Karen Coyle: great, tom. fortunately i had a copy

Thomas Baker: I had archived it when we deactivated the moinmoin wikis.