Data driven category based on structured tags - possible, how if yes?

[Bolitho]

Assume images with filled structured artwork tags.

(How) Can a data driven category be defined which delivers the following structure please:

creator one
- title of first work of creator one
- title of second work of creator one
...
creator two
- title of first work of creator two
- title of second work of creator two
...

etc.

I want to e.g. click on a creator to see all images containing artwork of this creator or
click on an artwork title to see all images showing this specific work.

Today I use categories which I assign manually (since structured tags were not supported before IMatch 2023). That does the trick.

I'm thinking over replacing these categories with the artwork tags plus a respective data driven category.

[Mario]

Would not a data-driven category with two levels do the job?
First level Artwork.Creator, second Level Artwork.Title?
This produces one category per creator. And below that a category for each artwork of that creator.

[Bolitho]

Doesn't work if you have multiple artwork from different creators in one image.
The same artwork is "assigned" to all creators in this image following your straight forward suggestion.
Or what is the specific configuration "trick" to avoid this?

[Mario]

Doesn't work if you have multiple artwork from different creators in one image
 The same artwork is "assigned" to all creators in this image following your straight forward suggestion.

This is what I would expect. If an image contains art from three creators, it should be categorized for all three creators.
IMatch cannot determine a preferred creator or something like that by its own.

If you have an image with artwork from 3 creators but you want to categorize it only for one of the creators, you have to come up with some concept like a preferred creator, stored in an additional tag or Attribute to make this info accessible for a data-driven category. Or you continue to categorize your images manually.

[Bolitho]

I'm confused - isn't that what structured tags are for?
I.e. having an entry for each piece of artwork (complete with title, creator etc.) in the artwork structure of an image?

Or may be I have been misunderstood when I used the term "assigned":
The problem is: The data driven category (as suggested by you) shows the title of each artwork under each of the creators in one image such as

creator one
- title of first work of creator one
- title of second work of creator one
- title of first work of creator two
- title of second work of creator two

creator two
- title of first work of creator one
- title of second work of creator one
- title of first work of creator two
- title of second work of creator two
...

[Mario]

Data-driven categories work that way.
On the first level, they create on child category for each unique value in the creator tag.
On the second level, for all files with the creator on the first level, a child category for each unique title is created. Since you have multiple creators with multiple titles for the same file, this is expected result.

[Mario]

Show me an example of the artwork metadata you have added to a file.
I assume the file is linked to multiple creators and shows multiple works of art with multiple tiles?

[Bolitho]

Just installed and tested IM2023.1.18

So this seems to be a bug (or not yet implemented).

Data driven categories don't work on the individual entries/lines of structured tags.

I.e. if you have e.g. an image showing multiple artwork of various artists and you add a "line" for each piece of work to the structured artwork tag a data driven category based on this structured tag doesn't distinguish between the different lines.

To reproduce it's sufficient to use one image only and add several different lines to the structured e.g. artwork tag.

A two level data driven category based on two elements of this structured tag wrongly shows all combinations of contained lines (as already described in the earlier posts).

[Mario]

This is how data-driven categories work:
A data-driven category looks at each distinct value for the used tag or variable, produces one child category for each distinct value and assigns each file with the same value to that category.

They have no concept for or knowledge of XMP tag sets or structures. They work on  individual tag values per level.


I add ArtworkOrObject|Creator to a set of files, say C1, C2, C3, C4, C5.
For some fies, I add multiple creators, e.g. C1 and C3.

A data-driven category on the ArtworkOrObject|Creator tag produces 5 child categories: C1, C2, C3, C4, C5 and each category contains the files of the corresponding creator. A file which has creators C1 and C3 will thus show in both the C1 and the C3 category.

If you add a second level, e.g. for ArtworkOrObject|Title, the data-driven category again selects all distinct values for that tag, and for each parent category (C1, C2, ... C5) it creates child categories for each Artwork Title where the creator matches the parent category.

If creator C1 has titles T1 and T2, this produces

C1
  |- T1
  |- t2

If you have an image with two AorO entries, e.g. C1 with T3 and C3 with T4, you'll also get

C1
  |- T1
  |- T2
  |- T3 <- Shows images of C1 and C3
...
C3
 |- ...
 |- T4 <- Shows images of C3 and C1

This is how data-driven categories work.
I don't know what you expect otherwise, but maybe not even IMatch can do everything.

[Bolitho]

What I would love to see is artwork (title) listed according to the respective creator.

I admit I expected the challenge if the data is stored in a repeatable! structured tag like artwork - this is why I raised the question.

It would really bring IMatch to the next level if it could distinguish between the different lines of structured tags and provide means to produce data driven categories accordingly.

[Mario]

What I would love to see is artwork (title) listed according to the respective creator.

It does. But if you mix multiple creators in the same file, the data-driven category must reflect that.
It is not build on creators or titles, but on the metadata contained in files.

Doping what you probably want to do is not possible with data-driven categories. Or iit would require some really substantial (and expensive) changes. Data-driven categories process one tag or variable per level. They have no idea about complex interrelations between individual elements in structured tags or need to.

Feel free to add a detailed  feature request. If a substantial number of users would say they need this, I will look into it for some later IMatch versions.