User:Jneubert/PM20 modelling

MOVED TO Wikidata:WikiProject 20th Century Press Archives/Data structure

Overall model

1. attach PM20 folder ID (P4293) to the item directly correspondig to the real world thing
   • persons
   • companies (not always matching exactly, but probably close enough)
   • sometimes wares

   e.g., petroleum in the United States (Q7178964)

2. create new item for the folder
   • instance of (P31) -> PM20 country/subject folder (Q91257459) <- subclass of (P279) <- PM20 folder (Q91257126)
   • this is extensible to new folders discovered later (not present at a frozen PM20 site)
   • as subclass of collection or creative work? or both? -> subclass of dossier (Q19515368)
   • NOT using archival collection (Q9388534) or special collections (Q4431094) - these are collections of folders
   • connceted to subject via main subject (P921)
3. new property for lists of folders
   • by geo
   • by subject
   • by ware

   for connection to real world item (sometimes close match)

Subject categories, wares, and geographical locations

• Countries and wares probably can be mapped to real world items, subjects are too arbitrary (e.g., "Land und Leute, Politik und Wirtschaft, Allgemein | Country and people, politics and economy, general)", "Postwesen. Telegraphenwesen und Fernsprechwesen | Postal services, telegraphy and telephony)" (and subcategories), "Geschichtliche Vorgänge 1900-1914 | Historical events 1900-1914")
• Subject categories must be represented in Wikidata as item for class, for
  • normalization (functional dependency of category name and notation from id)
  • representation of the category hierarchy
  • extensibility, when new folders should be added from films (versus a frozen HTML version of the classification at a certain point in time)
• BUT: the resulting system with facets from real world and from "PM20 items" defining a folder may look strange

Draft: Subject folder items defined by PM20 classification + existing location

New properties

Open question: Use ID (precise, but closed) or notation (more fuzzy, extendable) Probably: use notation/signature/code, for the reason given above.

name	pid	datatype	links to	comment	temporary use for example creation
PM20 subject ID		external	list of folders by country	at PM20 class item	catalog code (P528)
PM20 location ID		external	list of folders by subject/ware	at real world location item - makes sense?	postal code (P281)
PM20 ware ID		external	list of folders by country	at real world ware/product/product class item - makes sense?	-

Alternatives:

• one generic property for notation (something like skos:notation - already existing? preliminarily use short name (P1813)?), similar to catalog code (P528), in combination with catalog (P972) No - properties require different scopes and different formatter URLs
• Or: Use with formatter url instead of list property? No - only works as a non-extensabke list (generated HTML page) + lookup mechanism (notation -> id). Preliminiary implementaion with Skosmos leaves the list hidden as rdfs:seeAlso.

PM20 subject category (Q92707903)

currently ca. 1400 classes

example items:

PM20 subject category system (Q92732036)

e Health situation, general (Q92714111)

e1 Individual diseases and their control (Q92707235)

type	property	pid	datatype	cardinality	source property	transformation
Lde					label
Len					label	to add manually
Dde						"Systematikstelle des Pressearchiv 20. Jahrhundert" (fix)
Den						"Subject category of the 20th Century press archive" (fix)
P	subclass of	P279	item	1.1		super_class()
P	PM20 subject ID	Pnnn	external	1.1
P	main subject	P921	item	0.1		manual lookup

type: L=label, D=description, P=property, I=implied property

Alternatives

• Hierarchy could be represented as not as subclass relations, but as part_of/has_part hierarchy. Two possible advantages:
  1. no ontologically suspect is-a relationships
  2. enumeration of all parts in the super category

PM20 country/subject folder (Q91257459)

currently ca. 9000 subject folders

example item: Germany : Individual diseases and their control (Q91257808)

type	property	pid	datatype	cardinality	restriction	source property	transformation
Lde						skos:prefLabel
Len							derived from English location and class labels?
Dde					"Mappe aus dem Pressearchiv 20. Jahrhundert" (fix)
Den					"folder of the 20th Century press archives" (fix)
I	instance of	P31	item	1.1	PM20 country/subject folder (Q91257459) (fix)
P	location	P276	item	1.1	subclass/instance of human-geographic territorial entity (Q15642541)	zbwext:country	lookup_country() - may be country, group of countries, geographical region, ...
P	facet of	P1269	item	1.1	instance of PM20 subject category (Q92707903)	zbwext:subject	lookup_pm20_class()
P	main subject	P921	item	0.1			lookup_subject() derived from PM20 class
P	IIIF manifest	P6081	url	1.1			manifest_url() TODO
P	PM20 folder ID	P4293	external	1.1	starting wth "sh/"	dct:identifier

Ware folders

Same way as subject folders?

Company folders

In case of further work with company films, are folders needed?

Full representation in WD

• Folders with only one free document (as of 2018-12-28)
  • wa: 831 of 2891
  • sh: 1926 of 8842