Assigning "themes" to the visual content of an object?

[beaudet]

Hi,

I'm trying to avoid using metatypes outside of my Type definitions, Given that constraint, is below a reasonable approach for assigning multiple terms to the visual content of an object?

Should "Theme Assignment" be minted as a new term rather than using the existing term "Theme" for classifying the Attribute Assignment or is it reasonable to assume the metatype of an Attribute Assignment would define the metatype of the property being assigned assuming there's no other property like assigned_metatype that can be used for that purpose?

Finally, while redundant, it seems like a good idea to include both structures in the data, thinking of the attribute assignment as being layered on top of the basic represents assertion.

Thanks,

Dave

{
  "@context": "https://linked.art/ns/v1/linked-art.json",
  "id": "https://example.org/artwork/1234",
  "type": "HumanMadeObject",
  "_label": "A painting depicting a French War scene",
  "shows": [
    {
      "type": "VisualItem",
      "_label": "Visual content relating to 'French' and 'War'",

      "represents": [
        {
          "id": "https://example.org/type/french",
          "type": "Type",
          "_label": "French"
        },
        {
          "id": "https://example.org/type/war",
          "type": "Type",
          "_label": "War"
        }
      ],

      "attributed_by": [
        {
          "type": "AttributeAssignment",
          "classified_as": [
            {
              "id": "https://vocab.example.org/activity_type#School",
              "type": "Type",
              "_label": "School"
            }
          ],
          "assigned_property": "represents",
          "assigned": {
            "id": "https://example.org/type/french",
            "type": "Type",
            "_label": "French"
          },
          "assigned_label": "French"
        },
        {
          "type": "AttributeAssignment",
          "classified_as": [
            {
              "id": "https://vocab.example.org/activity_type#Theme",
              "type": "Type",
              "_label": "Theme"
            }
          ],
          "assigned_property": "represents",
          "assigned": {
            "id": "https://example.org/type/war",
            "type": "Type",
            "_label": "War"
          },
          "assigned_label": "War"
        }
      ]
    }
  ]
}

[beaudet]

Is school really part of attribution? Clarify first to make sure it applies here and not in production.

[beaudet]

from chat on call, edward suggested this might be most accurate:

VisualItem - created by - Creation - influenced by - Group - classified as - School, French

Not a lovely pattern, this.

[kkdavis14]

Right, (physical) objects are produced, (conceptual) works are created--do you mean what is the semantic difference between a production and a creation or what is the semantic difference between work and object?

[beaudet]

I think you answered my question. I was wondering why you're indicating any create/produce activities under the visual item, but it makes sense if you're modeling conceptual objects. Is it possible to find those easily through the LUX front-end? I'll probably need to study some of those for a future iteration of our linked art.

[beaudet]

This is probably a bit out of scope, but did the LUX team have UI discussions about how to treat conceptual works vs. physical objects?

[kkdavis14]

Yes, there have been many (ongoing) discussions about this. Right now it's handled by the block at the top of Object and Work pages that connects physical things and the conceptual things they carry--our language currently is "This object includes the following works" and "This work is included in the following objects".

Here's a Visual Work with a pretty fleshed out Creation Activity:
https://lux.collections.yale.edu/view/visual/e83fb78a-a767-4f0c-845e-506c78ff7c9e

Though truthfully, the modeling on that record is...not amazing. Room for improvement, definitely.

[beaudet]

Thanks so much!

[beaudet]

Definition of school within the context of our collection data:

When applied to art objects:
"The geographical and cultural milieu with which an object is associated. Although school is generally considered descriptive of the artist (see the column labeled School (Person)), it is maintained at the object level to accommodate those instances in which the artist is unknown and/or the work is associated with more than one school."

When applied to artists:
When the person is an artist, this column identifies the geographical and cultural milieu with which the artist is associated. Although an artist's school and nationality may be the same, nationality is determined solely on the basis of place of birth or country of legal citizenship, whereas the school classification takes into account not only the origin of the artist but the origin of the artist's style and the place(s) where the artist flourished. School reflects historic rather than modern geographic and political boundaries.

Examples
American
Antwerp
Flemish
Venetian
Italian

It sounds like semantically, the data is best associated with the creator and not the object so I'll see if we can just remove the school term from objects since it's problematic.

[beaudet]

My notes from the 1/22/25 linked art call about this:

represents might not be the right way to show this as a school -

it's stylistic - which means a metatype of style

but cat clearly applies because it represents cats -

all of them roll up into classified_as - can throw types at the visual items to apply classifications

so the keywords would be collected and brought together from those buckets -

keeping them separated from humans on the front-end is important - different types of keywords

rob recommends dividing them up a little more into classified, represents, and represents instance of

George: thematic vs. objective classification - use attribute assignment for keywording who applied it when, etc. for the context. The other option is to type another metatype

Rob - important to distinguish bank "as in river" from "bank" as in cash - applied more to string data rather than terms from a taxonomy, although that's a good topic for more discussion since taxonomies don't always clearly imply a metatype themselves and the metatype used in applying the term can be different from the metatype a taxonomy would naturally have for that term.