Closed edwbaker closed 1 month ago
I've been thinking some about this issue again in the context of the severe need for people to describe that an image or a ROI is a label. #200
Despite the various potential problems related to whether our use of Iptc4xmpExt:CVterm
is exactly correct based on the IPTC specs, it has been a part of AC for a long time and from the standpoint of stability, it might be best to just keep it rather than coming up with a new term or recommending something else.
I was thinking a bit about how I've seen dcterms:subject
being used in other contexts. It seems like it is more commonly used to describe the topic of something than to describe the actual thing itself. For example, a photo of people carrying signs in a protest might have the subject as "protest", whereas an actual description of what's in the photo would be things like "person", "sign", "flag", "fire", etc. In the library realm, one would probably use Library of Congress subject headings as values, and again it would be more what a book or images is "about" rather than what's depicted in it. So I'm thinking that dcterms:subject isn't the best thing to use in this case where we are actually trying to describe what's in the image.
It seems to me that a reasonable course of action that would mostly be consistent with existing use but also promote clarity would be to do the following:
dcterms:description
be used for a textual description. This could be free text or a controlled value string if people preferred.Iptc4xmpExt:CVterm
be used with an unabbreviated IRI. That's allowable under the current AC guidelines and (probably) consistent with the intent of the IPTC (although they seem to suggest using a namespace abbreviation rather than a full URI). subjectPart
and subjectOrientation
. This would be a bit at odds with the existing pattern of x
and xLiteral
, but I think it makes more sense to use the terms we already have.
What I've suggested is consistent with the examples illustrated in https://github.com/tdwg/ac/issues/200#issuecomment-963567163
The larger project suggested in https://github.com/tdwg/camtrap-dp/issues/191#issuecomment-1030732612 would still be a good idea, but I can't see myself actually doing it at any point in the near future.
There would of course be people using Iptc4xmpExt:CVterm
with controlled strings rather than IRIs, since that's what's suggested in the existing AC guidance for the term. But moving people to IRIs would probably be best in the long run since controlled strings from a bunch of different random vocabularies aren't guaranteed to be unique.
Capturing some thoughts from Katja Seltmann:
Will be closed by #246
Following on from discussions at https://github.com/tdwg/camtrap-dp/issues/191
See also early discussion of potential problem
Various possibilities to improve the situation here: