MODS and RDF Call 2015-10-05

Time: 9am PDT / Noon EDT

Call-In Info: 712-775-7035 (Access Code: 960009)

Homework Reminder: 

Moderator: Steven Anderson (Boston Public Library)

Primary Notetaker: TBA (etherpad for roll call: https://etherpad.wikimedia.org/p/RDF-MODS-20151005)

Attendees:

Agenda:

  1. Fedora 4 Test Code:
    1. Code base to try MODS title element in Fedora 4 -- hopefully will happen this week so people can take a look at how that's being represented within the next week or so.
       
  2. Overview of MODS name mapping attempts
    1. Anybody want to go over their mapping attempt who wasn't on the call last time?
      1. Julie: missed last week's call, has an IU name mapping attempt to go over
        1. Basically just used Dublin Core terms for her example.
      2. another option is to use MARC relator as URI instead of DC terms as Creator -- these are sub-properties of DC terms already, and are more precise than just Creator/Contributor

      3. DC:Creator and DC: Contributor with a literal text string; this actually has to be a URI value, literal values are not allowed

        1. Can literals be used with MARC relators? No, as sub-properties of DC:Creator and Contributor, they inherit the same restrictions

        2. how do you solve this? make your own URI, create a blank node?

    2. Document that talks about DC terms elements -- will be added to the notes afterwards, good reference for what can and can't be used with DC properties
      1. https://code.google.com/p/tdwg-rdf/wiki/DublinCore#1._Use_of_Dublin_Core_terms_in_RDF

  3. Favorite mapping that anyone wants to talk about that wasn't their own mapping?
    1. Steven: really liked the NYPL's idea of using a local copy of the prefLabel, can add additional properties and then if the linked data source ever goes down, can still display it in the UI relatively easily and always know that it has at least the prefLabel property.
      1. Version 2 of BPL's Document tries to implement this approach: https://docs.google.com/spreadsheets/d/1nNnGI-u9RazlFJ_cdDJPT6dpASjnqHwomGbpScDIATQ/edit?usp=sharing
      2. Eben: doesn't love the idea of creating a local name for something that already has its own URI, BPL example where the name is something that has an LoC URI but the example uses a local value, which seems counter-intuitive -- why create local names for things that exist in other places? Get the concept of having the text string locally available, but it seems redundant. Can someone from New York talk more about their rationale?
        1. Shawn: Not everyone is going to have the same use case as at NYPL where there is a requirement to store local data about names, whether local or already have a URI; are in a position where they may need to be publishing a dataset for a curator. NYPL's use may be overkill for many implementations. If you still wanted to have a similar use case you could do blank nodes, but not sure how well that would work. 
        2. Steven: lots of discussions on this over at BPL, if people need to add additional properties, concern is that when you start using linked data sources for locally minted names from other smaller organizations, higher probability that an end point may go down or that the institution may decide to no longer support that linked data set. By having a prefLabel, you have that for long-term preservation so you know what that URI was at one point and at least know what the name was. 
        3. Eben: not that different from what we currently do in MODS where the string is stored as a text node and have the URI as it exists in the URI value attribute on that name, just wondering if other people felt the same way that it seemed redundant. Philosophical question: where do you draw the line in terms of what URIs you're willing to rely on vs. what URIs you make local copies of the prefLabel for? How far does one want to take that going forward? 
        4. Julie: are you getting rid of your MODS once you've done this, or does the MODS XML hang around somewhere you have access to?
        5. Steven: this will replace MODS at BPL, maintaining both linked data and updated XML would be too much to handle.
        6. Julie: If you're dealing with both string and URI in XML, makes sense that you'd want to carry that forward. At Indiana, we identify if information is from a URI and where, but we don't store that information -- just keep the string value -- so coming from the literal route regardless.
          1. Steven: You could use this approach for one where you don't have URI support, could still have a resolution that would still have the prefLabel without the SKOS exactMatch property. Could still use the data model, just wouldn't be able to link out to another authority. 
          2. Julie: Makes sense, since Indiana doesn't have that info.

    2. Steven: if minting your own URI, something that's only relevant in your local area or etc., do you want additional properties that would be associated with that entry? 
      1. For ex: storing FOAF name, which might be different from the prefLabel in that that could just be name part, name w/date, best representation of base name, whatever. To follow that up, if a local-created name doesn't have an external URI, is the minimum that it would be just a prefLabel or does it have other properties we could expect? 
        1. This is especially as prefLabel will have to be updated from some kind of sidecar system every once in a while for external sources. So locally minted names could use a similar process to update their prefLabel where FOAF name could be canonical version of name and prefLabel could be generated from that and updated as properties change; if death date is updated, then index circuit could update prefLabel from FOAF name to have the correct death date at the time. 
      2. Question: are there other pieces of data besides birth and death date that get displayed with name? Terms of address, etc.? 
        1. Answer: titles, sometimes occupations, there can be quite a bit more; sometimes it's dates when person was most prolific, could be a role for disambiguation purposes 
      3. Question again: If people are only interested in the way that the name is normally displayed, are people fine with that being the only representation of the name that they're storing, or if they want another representation that's the canonical name without its accompanying dates or terms of address, do people have a use case where those things would need to be separated?
        1. Danny: Name is supposed to be a thing on which information can be colocated, so breaking it up into component parts makes colocation harder. John Adams and John Quincy Adams w/out dates makes it hard to facet on, not really in favor of breaking it up. Birth and death dates broken out might be good for searching on those dates, but to this point, the way these headings have been used, making them smaller than the full heading is not useful.
          1. Steven: can just try what seems to be best when they start the collaboration document
      4. Shawn: Do we need to make a recommendation on how we store the names behind the URI, or do we really need to just make a recommendation on having a URI/string/option for both? 
        1. Steven: if we write a shared code base to translate from MODS XML into whatever MODS RDF ends up being, it's necessary to make some of these decisions for what it spits out. So mostly just in terms of collaborating coding efforts more than whether or not it's syntactically valid

        2. Remark: so much variation in how people are structuring name or using it, how general is this going to be?

        3. There's a lot of particular use cases, people are always still welcome to do the things that they want to keep doing, nothing is binding, but this is a long process and there is something to be gained from having some kind of consensus; still, "recommendation" has to be taken with as many grains of salt as possible based on use case.
        4. Julie: we're turning what we're doing into an actual transformation, can we look at this more as creating a base transformation for people to go from? Can t