NPGConstituents: Name fields

[steads]

The use of E82 Actor Appellation is incorrect. These are instances of E41 Appellation applied to instances of E39 Actor NOT instances of E82 Actor Appellation as they are not characteristically of a form that indicates that they are being applied to an instance of E39 Actor. The use of P106 is composed of (forms part of) to decompose the appellations into constituent parts is correct. However many of the constituent parts of names are not Appellations but descriptions of group membership. Groups need to be separated from individuals and then the true meaning of component parts and the relationships to each other and the instance of E39 Actor can be assessed. Indirect naming; daughters of; man of x family etc do not show the genuine nature of the real world objects so named. => More sophisticated rules and mapping preferable. DisplayName, NameTitle, FirstName, MiddleName, LastName, Suffix, AlphaSort and Institution are probably all appellations and can be properly mapped once the mixture of E21 Person and E74 Groups is resolved and the various membership relationships have been become clearer. There could also be P1 links between the constituent appellation parts and E39 or E21. This would make some kinds of integration easier.

[VladimirAlexiev]

• I think the use of P131_is_identified_by and E82_Actor_Appellation is correct, not sure what is the issue
• I checked the breakdown of names into fields. There are many mistakes, eg

DisplayName	FirstName	MiddleName	LastName	NOTE
Ted Bair and Harvey Filister	Ted Bair and Harvey	NULL	Filister	Ted Bair is a full name, not first name
Tench & John Frazer	Tench & John	Frazer	NULL	Tench & John are two people
Teodoro de Croix	Teodoro	de	Croix	de Croix is the last name
Teresa J And Grace Fitch	Teresa J And Grace	NULL	Fitch	Teresa J and Grace are 2 people
Marianne Von Geyer Belcher	Marianne	Von Geyer	Belcher	She has two last names: Von Geyer (maiden) & Belcher (married). http://www.biographi.ca/en/bio/belcher_edward_10E.html
Wilhelm von Moll Berczy	Wilhelm	von Moll	Berczy	von Moll is a former last name, see https://en.wikipedia.org/wiki/William_Berczy
Gebhard Leberecht Von Blücher	Gebhard	Leberecht Von	Blucher	Von Blücher is the last name

• However, there are also many correct splits. So overall, I think the name fields should be kept. P106 is composed of is used correctly
• But as an optimization, name parts should not be emitted if there's a single part. Eg this below is superfluous:

<http://americanartcollaborative.org/npg/person-institution/0/appellation> a crm:E82_Actor_Appellation ;
    rdfs:label "Unidentified Artist" ;
    crm:P106_is_composed_of <http://americanartcollaborative.org/npg/person-institution/0/appellation/lastname> .

<http://americanartcollaborative.org/npg/person-institution/0/appellation/lastname> a crm:E82_Actor_Appellation ;
    rdfs:label "Unidentified Artist" ;
    crm:P2_has_type <http://americanartcollaborative.org/npg/thesauri/nametype/lastname> .

[steads]

Instances of E82 Actor Appellation must, by there very nature, be identifiable as appellations of actors rather than be appellations which are used to identify actors. Things like ULAN identifiers, social security numbers and company numbers are examples of E82.

You have correctly identified even more of the many problems in the splits!

There is a much bigger problem with the "Unidentified ...." in that the same "person" is used to stand in for all unidentified people!!

I agree that it is madness to create "last names" out of things which do not have such a concept like "unidentified..." and company names.

[VladimirAlexiev]

Hold on, "John Smith" is not an actor appellation??

Agree about Unidenfitied, maybe post a separate issue for it?

[steads]

No John Smith is not an E82 Actor Appellation

[edgartdata]

Can we come back to E82 versus E41? Is that a deal breaker?

Also I am not sure that modeling appellation parts (first name, middle name, last name) separately is useful? Unless someone researches naming trends I guess? Wouldn't it be more useful to have the literal/label with the actor's full name?

[edgartdata]

Can we work on a solution for unidentified actors? It will be important because otherwise we will leave out many actor entities out of the AAC browsing app.

[steads]

Each Unidentified Actor must have there own separate unique identity which all share the same instance of E41 Appellation {Unidentified}.

[edgartdata]

So E39 or E21 individual Unidentified Actor -> P1_is_identified_by -> E41 Appellation {Unidentified} where E39 or E21 individual Unidentified Actor get unique URIs?

For example one such actor may be http://collection.britishart.yale.edu/id/page/person-institution/185 and another one may be http://collection.britishart.yale.edu/id/page/person-institution/3887 (made up URIs).

The problem with that, at least at the YCBA, is that all individual unidentified actors are lumped all together into 1 TMS authority record (does not make sense in terms of semantics but is practical for managing traditional data pre RDF - otherwise our db would have many similar similar Unidentified Actor authority records). So should we put a process on the URIs for these Unidentified Actors that whenever they are linked to a different object they get a different URI? There must be a better solution!

[steads]

Can we come back to E82 versus E41? Is that a deal breaker?

Nope just use E41 it is the superclass for E82 anyway

[VladimirAlexiev]

Hi @steads! I still don't understand why 500010654 is more "identifiable as an appellation of an actor" (to me it looks like a mere number) than "Michelangelo Buonarroti" is.

@edgartdata: yes, strict modeling would require making unique "Unidentified" URLs per object. I think we need a separate issue for that?

[steads]

@VladimirAlexiev cannot help with your understanding!! Just telling you what the CRM definition says.

[workergnome]

@steads, I'm trying to understand where in the CRM definition that is. If I look at http://new.cidoc-crm.org/Entity/e82-actor-appellation/version-6.2.1, I see a list of examples which include things like "John Doe" or "International Council of Museums", but nothing that looks like a raw social security number (though a string including a social security number is included).

Is it possible that I'm looking in the wrong place?

[si-npg]

All of Alexiev's examples of "mistakes" in are constituents that should not be included in our data at all. It will help to get NPG's second export uploaded so that no time is wasted analyzing data that won't be included. I will contact Craig, who has the cleaner data in Dropbox.

[VladimirAlexiev]

Quickly checked NPG_LOD.zip. 12.5 records.

The parsing is much improved, eg

• the parsing of names with up to 3 component seems perfect. Applies to English names, but not only
• perfect parsing of Title (eg "Captain", "Mr", "Ms") and suffix (eg "Jr")
• Chinese names are inverted, eg "Zhu Rongji" parsed as first="Rongji", last="Zhu"
• most Institutions seem to be properly identified
• most nicks are properly handled, eg Henry "Hank" Aaron, Arnold "Red" Auerbach

Some problems:

• "John Andrew and Son" is parsed as a person with suffix="and Son". But eg "Henry Bryan Hall and Sons" is parsed as Institution
• some problems with multiple people, eg "Thomas C. and David C. Collins" is parsed as first="Thomas C and David C"
• a few Institutions also have first/last, eg "Daniel Rice and James G. Clark" is parsed as first="Daniel", last="Rice" (James Clark is skipped)
• Spanish names may present a problem. Eg "Ernesto Zedillo Ponce de León" is commonly known as "Ernesto Zedillo", eg see https://en.wikipedia.org/wiki/Ernesto_Zedillo. His father is "Rodolfo Zedillo Castillo" and his mother is "Martha Alicia Ponce de León". The data says last="Zedillo Ponce de Leon" but I'd say last="Zedillo" or that he has two last names.
• Sometimes middle initial is stripped of fullstop (eg "Henry T. Anthony"), other times not (eg "Charles S. Abbe")
• some nicks are treated as middle name, eg William A. "Bud" Abbott, Marjorie "Marge" Champion, Nat "King" Cole, Louis Francis Cristillo "Lou" Costello
• some titles are parsed correctly, but eg "Charlotte-Marguerite de Bourbon, Comtesse de Lowendahl" is not: first="Charlotte-Marguerite de Bourbon", last="Comtesse de Lowendahl"