How to organize people changing names (best practice)?

[akirot]

IMatch 2020 is great, thank you for this new release!

Nevertheless I need some advice, please.
Face detection and recognition work as expected and I studied the comprehensive help - the new features are overwhelming.
Now I'm lost: How does one handle changing names (e.g. by marriage) of a person?

Currently I use a simple category structure:
Initially a specific person on an image is assigned to a category "FirstName LastName". As soon as this person marries I implement an additional category "FirstName NewLastName (née Lastname)" resulting in a hierarchy like this:

FirstName NewLastName (née Lastname)
- FirstName LastName

The person is assigned to the new category from now on.
Thus I can select easily the images of the very same person at various stages of life.

My first idea was to continue with this approach and use the feature to assign categories to people and change the assigned category as soon as the name changes. But this can't work since (according to the help) this change would affect all already tagged images, not only the ones after the change.

Any recommendation? How do you do that?

[Jingo]

Not sure it is the best way - but I would probably just edit the person, change the Full Name to the new name and add the old name to the Other Names field. 

[akirot]

@Jingo: Yes, this is an approach pointing into the right direction. The drawback I see: As soon as you touch an "old" image containing this person (e.g. change keywords), the related information is updated - resulting in "wrong" metadata for the "old" images.

Just reread the Help. One can clone a person. I would be fine if I had more than one entry for the same person (e.g. one for the unmarried and one for the married person).
The drawback I expect: The digital fingerprint of the two people entries would be identical, the face recognizing AI can't decide any longer which person is the right one.

Brainstorming: If a people entry contained a flag like "aktive/non active" for face recognition one could tell the AI which entry to use (or to ignore) - whilst this is a straight forward approach it would require a feature request. I'm happy to raise this - but probably there is a simpler approach?

[Mario]

The other names property of persons is designed to keep birth names and other 'previous' names.
You can use these as the label or access them via variables if required.

[Jingo]

@Jingo: Yes, this is an approach pointing into the right direction. The drawback I see: As soon as you touch an "old" image containing this person (e.g. change keywords), the related information is updated - resulting in "wrong" metadata for the "old" images.

Hmm.. which metadata would then be marked as wrong?  For keywords, I keep a combination of old name / new name via synonyms... ie: Betty Sue Mellon is also known as Betty Sun Veggie (married name).  Best of both worlds as you can locate the person by either tag.  This can also be accomplished via attributes, labels, categories...

[akirot]

@Jingo: Probably I'm mistaken.
Taking your example, assume Betty Sun Veggie at a certain point of time is added as a synonym (or additional keyword, category ... ) to the people entry of Betty Sue Mellon. As soon as you touch an old image of Betty Sue Mellon it will get the Betty Sun Veggie synonym, too (because it is now configured in the people entry). This is wrong since she was Betty Sue Mellon "only" at the time the old! image was taken. (I understand always the then current people entry is written back to an image if there is metadata to write back - or do I get this wrong?)
My current approach (category hierarchy, newest name on the top level, oldest name at the bottom level) addresses that.

[Mario]

I can't follow, sorry.

And IMatch and the People feature cannot handle all possible fringe cases or exceptions.
Doing so would over-complicate things for all other users. We have enough of that in IMatch already.

To keep previous names associated with a person, use the "other names" property.
If you run into such problems often, try to keep this kind of volatile data outside of person metadata. Use XMP metadata, keywords, categories, or Attributes to handle this type of data. IMatch offers you many ways to do this.