Discussion: are AI labels the end of hierarchical keywords?

[Uwe]

Hello,
I have not installed the beta version 2025, but I would like to discuss a point that I noticed when looking at the AI labels. I have registered the same thing in other programms that generate tags "automatically": The tags are flat - i.e. no hierarchical structure. Does this possibly mean the end of hierarchical tags/keywords? Am I looking too black?
What is your opinion on this?
regards, Uwe

[axel.hennig]

Where did you find/see the "beta version 2025"? I don't think that Mario is providing something like this.

[Mario]

Nah, IMatch has you covered!

A flat list of keywords is rarely helpful, unless you get by with maybe 50 to 100 keywords, tops.
I'm just spending evenings and nights writing the help for the new IMatch AutoTagger that can utilize AI in the cloud and running locally. The results beat everything we've had before in precision and quality. This will be good. Especially when the unique IMatch Traits concept comes into play.

IMatch AutoTagger offers several features to optionally "process" the flat keywords produced by AI.

Thesaurus

AutoTagger can map keywords via your thesaurus into your keyword hierarchy.
For example, if the AI delivers the keyword beach, and your thesaurus has the hierarchical keywords Nature & Environment|beach and Travel & Geography|beach, AutoTagger assigns both hierarchical keywords to the image.

AutoTagger searches keywords and synonyms to find matches for flat keywords delivered by AI. So it will be able to map both "beach" and "beaches" to your hierarchy when the keyword "beach" has the synonym "beaches".

Manual Mapping

This is a feature that allows you to map incoming keywords delivered by the AI to zero or more outgoing keywords.

This feature can be used to blacklist (ignore) keywords you don't want to import.
It can be used to fold variations of keywords into a standardized form, e.g beach, beaches => beach.
It can be used to extend incoming keywords into variations: beach => beach, beaches
It can be used to translate keywords delivered by the AI: beach => Strand or beach => beach, beaches, Strand, Stände



Of course all of this is optional. You can just use the keywords delivered by the AI and AutoTagger 'as-is'.

IMatch's AI Tags

This is an important addition to IMatch. IMatch 2025 offers dedicated "AI Tags" for keywords, descriptions, landmarks and traits. This means you can use the AI results in addition to keywords and descriptions you enter manually, without polluting the human-made data with AI data.

If you use this, you have basically two sets of keywords and description for an image. The description and keywords produced by the AI create a much richer search experience, because you can find files by keywords or words in the description you would never have thought of adding yourself.

Create a data-driven category for the new AI.keywords tag automatically organizes your files based on the keywords the AI has produced.

(Btw. there is no beta version of IMatch in the public. I just discuss openly in posts what's brewing in my lab for IMatch 2025).