minutes20150813

DCMI RDF AP Task Group
http://wiki.dublincore.org/index.php/RDF_Application_Profiles
Meeting date: Aug 14, 2015
Meeting link: https://meetings.webex.com/collabs/#/meetings/detail?uuid=M1RMXAWHCVXID8E5NEAFH8K88R-JV0D&rnd=233032.37264
Meeting minutes: https://etherpad.wikimedia.org/p/dcmi-ap-13-08-2015

Attendees: Karen, Corey, Antoine, Hugo
Regrets:  Valentine
    
AGENDA:
    
1.   Deliverables for publication

 a. Use Cases
 http://wiki.dublincore.org/index.php/RDF_Application_Profiles/UCR_Deliverable
 b. Requirements
 http://wiki.dublincore.org/index.php/RDF_Application_Profiles/Requirements       

ACTION: Karen to complete header and write introduction; link to database/heading in UCR document
 Say that reqs us a supplement to the UCR document
 DONE

ACTION: Karen to coordinate with Stuart on announcement - probably needs
 something written up, esp. regarding coordination with W3C
 DONE: Stuart is thinking that they should be listed on the "recommended
 resources page on DCMI web site:
     http://dublincore.org/documents/#recommendedresources
 
 
 2. Requirement database
Changing names and URLs:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=DC-ARCHITECTURE;77251466.1507
DONE
Creating dumps:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=DC-ARCHITECTURE;65efdc4c.1507
DONE - dumps can be created
coordinate dump and this version of the documents
ACTION: Antoine to create dumps and link them from the wiki pages
ACTION: Thomas change PURL redirect


3. Planning for AP development   
    What are the next steps to develop an AP? 
    
    We have Singapore Framework, DCAM and DSP.
    
Note: There are differences between DCAM and RDF (esp. treatment of SES as non-literal). Should we try to follow RDF/OWL in our basics
rather than DCAM?

We have to branch away from DCAM. RDF is our focus
the Singapore framework is still usable
DCAM and DSP can be seen as a 'version 1'
DSP is roughly what SHACL has already done
But there are still important concepts that are not touched in RDF/SHACL
documentation aspect of DSP is important for us
 - for humans
- for machines (actionable)

Defining the graph -- tools in the wild already do this.
 - do we have enough common practice to define this?
 - can we gather examples?
 There are still many things to discuss.
 
 Karen: I need to chat with Arnaud and Eric. I will send them an email.
 
 There's a rift on these topcis between the two groups, but it may not be representative at all of the respective position of us (focusing on open data) and private sector. There might be not much difference here...

Question, whether these issues are at the validation level
One thing's sure. It's in scope for AP.

It could relate to the new LDP working group.
LDP is more for data exchange
But that's what the 'de-referencing' and 'graph definition' are, actually. Very much akin to ETL problems.

Corey: I can have the discussion in Minnesota, in the end of September.
Karen: it would be great to create a picture before then. A bit like the Singapore framework.

ACTION: Corey to record a picture

Do we need a connection with the next LDP\

Should we develop this right now?

Karen: we could be a group where we follow and discuss
... build requirements over time.
    
    

Look at Corey/Tom/Karen document and see how it fits into this in terms of defining the scope of an AP.
    http://wiki.dublincore.org/index.php/RDF_Application_Profiles/blendingDoc
    
[to be cleaned
    a. Do we need new/additional requirements?
    - as per Lars' question: AP needs to be machine-readable and allow you to process the data.
    - graph selection is an issue
    - DSP section #3: Binding of statements to statement templatesFor each description, each statement is bound to a Statement Templatein the corresponding Description Template by evaluating the Property Constraint.Each statement must be bound to exactly one Statement Template. 
    - Corey: validation is a serialization concern; 
    - How do you figure out which profile to apply even if you don't have classes that defines a specific shacl validation
    step?
    It's possible that the application needs to do that, that it 
    won't be in shacl.
    - this is what an AP can provide: information about the data or data source that is not
    defined as an RDF classes
    - possible aspects of validation: by source; by material type; by some other aspect of the data
    
    SHACL could be used today for EDM data, but for data as complex as is found in MARC, more decisions
    would be needed outside of SHACL on what data gets what kind of validation.
    These questions could be answered with application profiles. AP would parcel out data to SHACL templates or engines.
    
    Latest shacl ttl file? https://raw.githubusercontent.com/TopQuadrant/shacl/master/src/main/resources/etc/shacl.shacl.ttl
    example: https://github.com/TopQuadrant/shacl/blob/master/src/test/resources/shaclsquare.shacl.ttl
    
    b. Do we need a discussion of how the DCAM fits our needs?
to be cleaned]

ACTION: Antoine to look at unclear requirements in Hugo's email
--ONGOING: only one requirement remains unclear: https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=DC-ARCHITECTURE;8c92de00.1506

Lars' new requirement, Discussed in previous call: https://etherpad.wikimedia.org/p/dcmi-ap-18-06-2015
ACTION: Antoine to sum up and email the discussion. [on the differences with other requirements] And put the case and requirement in our database.



4. Alignment with W3C work

Recent discussion: re adding ability to check that IRI has certain attributes based
on dereferencing:
    starts:
        https://lists.w3.org/Archives/Public/public-data-shapes-wg/2015Aug/0007.html
   view thread:
    https://lists.w3.org/Archives/Public/public-data-shapes-wg/2015Aug/

http://www.w3.org/2014/data-shapes/wiki/Main_Page

Optimization requirements for boolean combinations: 
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=DC-ARCHITECTURE;e018b30.1508

SHACL and RDF: SHACL is written in SHACL and cannot be processed with
tools designed for RDF. Is this a problem for us?
Do we have something to say to the W3C group about this?

Email thread on Darwin Core, SHACL, and APs

Comparison of W3C and DCMI requirements:
https://docs.google.com/spreadsheets/d/1bCpQVyxI-N2Ca83umvQD8OKTdsDyG6Sz-E8Qo3v8ynM/

Hugo's email with suggestions: https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=DC-ARCHITECTURE;41aa27ca.1505
ACTION: group to tackle by email:
- requirements taken from W3C which apparently don’t have a match in DC
- suggestions on the mapping from DC to W3C


5. AOB

Scope
Validation?
Need for testing the SHACL technology
Hugo, Valenitne and Antoine are still interested in validating Eurpeana data

Karen: there is a list to test cases
... we could provide one

Corey: I'm willing to keep in the group, acting as liaison, but I'm not sure how much I can commit day-to-day


X. Express requirements as RDF, mapping between DSP and our requirements


X. Coordination with BIBFRAME
Thomas' evaluation of BIBFRAME AP
http://wiki.dublincore.org/index.php/RDF_Application_Profiles/BibframeAnalysis


X. Coordination with Hydra, Fedora
Hydra Metadata Working Group: https://wiki.duraspace.org/display/hydra/Hydra+Metadata+Working+Group
Portland Common Data Model: https://wiki.duraspace.org/display/FF/Portland+Common+Data+Model
ACTION: Karen: post this link to W3C group - on hold until W3C group gets to the right point


X. Coming conferences: DC, TPDL, SWIB


X. Glossary - postponed until we have time!