I completely agree.  I think that Adam and I were using the subfields as demonstrating the particularly egregious nature of bf:Title in straight translation from MARC without ( much / any? ) additional modeling work.  
The 100 vs 773 example is also a great point, as (I assume) both of those would become bf:Person.

In order to determine the right way forwards regarding bf:Title, having a set of use cases and requirements would be great.  There are clearly particular descriptive properties that are needed for a title, such as might populate the relationships between transliterations, translations, main vs sub vs part, etc. and having all of those in one place to assist with the decision would make that go much more easily.


On Fri, Jul 25, 2014 at 10:09 AM, Karen Coyle <[log in to unmask]> wrote:
Straight translations from MARC to [anything else] don't make sense unless you analyze not just the MARC coding but the intention and meaning behind the data element. (I covered this is in a code4lib journal article, to some extent.[1]) That some bit of data was broken up in a certain way in MARC is not necessarily meaningful in other situations. Title is a good example, and Thomale shows how hard it is to make sense based on MARC subfields.[2] One other example is personal author: the x00 field has a dozen or more subfields to encoding personal author data. In other MARC fields, all of that information is placed in a single subfield (because of the limit of 26 data subfields per tag).

100 $aHamilton, Milton W. $q(Milton Wheaton), $d1901-
773 $aHamilton, Milton W. (Milton Wheaton), 1901-

Yet one of the biggest problems in translating MARC to [anything else] is that because the data has only been expressed as text strings, no consistency checking has been done on the content. That 245 $p could have anything at all in it, including Aunt Martha's apple pie recipe.

Developing a future format based on this data is a formula for failure, I'm afraid. Yes, we will need to translate our old data to some new form, but that is not the same as developing a new format intended to mimic the data of old one.


[1] "MARC21 as Data: A Start"
[2] "Interpreting MARC: Where's the Bibliographic Data?"

On 7/25/14, 9:30 AM, Robert Sanderson wrote:

+1 to all of these :) And to add:

Do we need a title resource at all, or is a string sufficient?

On Fri, Jul 25, 2014 at 8:28 AM, [log in to unmask] <[log in to unmask]> wrote:
Having now for the first time taken a close look at bf:Title, I'm a bit taken aback.
1) "Title being addressed."

According to my deproliferation, I read this as "value".  And hence the actual string that's being turned into a resource.
If there's a need for a bf:Title (per 0) then there's a need for the value.

I think the question is, if there is a need, is this the entire title or only some part of it that's being decomposed into the resource.

Otherwise ... 

What is the purpose of this specialized property in the face of the bf:label that is available to all bf:Resources? What would cause someone to use it? Is this just MARC 2045$a in a new format?

Yes. Or is it $a plus $b?  Or does $b go in partTitle, but then where does $p go?  If we're recomposing, lets do it right :)

What about getting rid of bf:title and bf:titleStatement, and just using bf:label?  Quite independently of any other decision regarding titles as resources and their attributes.

2) "Qualifier of title information to make it unique."

Working for the uniqueness of labels goes very much against the practice of Linked Data. The Title entity is already possessed of an identifier. If anything more is needed to ensure uniqueness, isn't something badly wrong with the identifier?

+1.  As currently described, titleQualifier seems to do more harm than good.  There's no example, so it's hard to tell exactly what's going on.

Is this just 245 $v?

3) and

Is there any purpose to this distinction or is this just a case of MARC 245$n and $p being mechanically preserved? In fact these two properties have the same range.

+1.  If partNumber is really a number (which I don't think it is, as $n can hold "Part One") then it could be xsd:integer. Can a single title have both a partNumber and a partTitle?

3.5)  See question about $b

4) "Class or genre to which a Work or Instance belongs."
and "Other distinguishing characteristic of a work, such as version, etc.."

These seem very strange to me. In what way are these properties of a title at all? Is this just a mechanical transfer from MARC 245$k and $s? This seems to be information that should be recorded on the Work or Instance.

In order and in my opinion: No, Yes, Definitely :)

Same with copyright from Provider?  (Or better decompose provider back on to the Instance)


Karen Coyle
[log in to unmask]
m: 1-510-435-8234
skype: kcoylenet

Rob Sanderson
Technology Collaboration Facilitator
Digital Library Systems and Services
Stanford, CA 94305