duplicate agent weirdness matching on job title, and initials

[cjconroy]

Curious why Arctos is suggesting other agents are possibly duplicates only because they are also "Associate Professor" as in this case. Seems like a waste of Arctos's brain power.

[dustymc]

why

attribute_type + attribute_value == attribute_type + attribute_value

Associate Professor

yea, that's not very informative.

Arctos's brain power

It would take a bit more brain to add an "except when..." in there, but it's an irrelevant amount either way, we can do whatever is most likely to lead to places without duplicates, wherever that might be.

There are 933 distinct 'job title' values. A bunch of them are used about once and might actually help detect duplicates eg

Digitization Project Coordinator at the Field Museum of Natural History

the most-reused don't seem very useful for duplicate detection eg

Dr.                                                                                                            |    15
 student                                                                                                        |    17
 Three Seas Student 2025                                                                                        |    18
 Graduate Student                                                                                               |    18
 Collections Manager                                                                                            |    18
 Collection Manager                                                                                             |    24
 Student                                                                                                        |    27
 CMB Student                                                                                                    |    28
 Director                                                                                                       |    29
 Associate Professor                                                                                            |    30
 Curator                                                                                                        |    40
 Assistant Professor                                                                                            |    48
 Professor                                                                                                      |    78

... but those might be useful for other things, I'm not quite suggesting they don't belong.

There are a few that are probably misplaced eg

https://arctos.database.museum/agent.cfm?include_verbatim=false&attribute_type=job%20title&attribute_value==J.&include_bad_dup=false

The documentation is terrible/wrong

https://arctos.database.museum/info/ctDocumentation.cfm?table=ctagent_attribute_type#job_title

The agent's current job title.

(this could/should apply to any job/position/career/hobby/something - see below - and ideally there might be a history of these)

I'm also not quite sure we've got the term right ("potter" seems useful/intended here, but also suggests something slightly different than "job title.")

Can we change the "shape" of the concept to encourage more-useful data somehow?

Should this not be considered when duplicate-checking?

Bit of both, somehow, maybe??

[mkoo]

Should this not be considered when duplicate-checking?

I've argued before to drop in duplicate flagging so yes!

[Jegelewicz]

Please drop this, but also, just because an agent shares a first initial and last name. Today I was creating Tyson Burch and I had to add a relationship because Arctos tells me they might be the same person as Thomas Adams Burch - that seems a little nuts?

[dustymc]

nuts

You asked for that....

#438

[Jegelewicz]

So maybe I am nuts? Actually, I am OK with the "maybe it's this guy", but then needing to add some weird relationship hurts my brain and takes up time. Why can't I just force-create after deciding they aren't the same person?

[dustymc]

needing to add some weird relationship

If I've managed to figure out what you mean, that's not the only option (and not a terribly useful one in this case...).

I try to make an agent with no information at all:

... and get a hard block:

I back up and add ANYTHING....

... and then I can push them through.

This doesn't seem like an unreasonably high bar to me?

I made another with some really horrible "evidence" - FWIW I'd consider this vandalism and request a role review if I saw this in production so don't follow me....

and that's still all that's needed, ANYTHING, any little scrap of info....

& et voila... https://arctos-test.tacc.utexas.edu/agent/21364423

[Jegelewicz]

This doesn't seem like an unreasonably high bar to me?

So I just have to add a piece of garbage and it's OK? What you demonstrated is what I've been doing, then going and deleting the piece of garbage I had to add to create the agent. It's a waste of time. Why can't we be trusted to review the potential matches and then force create without making stuff up first?

[mkoo]

why are you adding in garbage? I force-create agents with joy if I can add a bad dup relationship! So please ignore Dusty's (presumably) facetious example!

[Jegelewicz]

if I can add a bad dup relationship!

For that one, it make sense, but I have had plenty where it doesn't.

[Jegelewicz]

For example. Attempting to create Ann R. Miller and get this

Who do I create the not the same relationship to? This person is clearly not the same as any of them or at least it is unclear if this person is. But I don't have the capacity to research every name on this list and as far as I can see the person I am creating is not any of the people in this list.

Those with real potential include:
A. Miller - which is missing the middle initial and based on activity could be almost anyone.
Annaliese Miller - who "Did illustrations for grass volume (V. 24) of FNA"

The Ann R. Miller I am creating collected diatoms, so I feel pretty confident they aren't any of the people listed. Sure, I can create a weird "not the same as" A. Miller - but WHY? I really don't have a reason to do that - who knows, maybe some of the activity associated with UAMb:Herb is actually Ann R. Miller and I am refuting that without any evidence. I can also try to get someone in that collection to confirm or deny, but how much work is required to add an agent?

[Jegelewicz]

How about this one. Attempting to create John Rudolph and get

Do I create this "not the same as" relationship? Why should I have to?

My intent in the issue linked above was to FIND potential dupes before creating an agent, not to prevent me from creating an agent that is clearly NOT a duplicate.

[campmlc]

Related issue with trying to create operators- ArctosDB/arctos#9314
Can we restrict the flags so that clearly inappriopriate matches - e.g. different first or middle names/initials are excluded?

[Jegelewicz]

They don't need to be excluded, they just shouldn't PREVENT force create.