Issue 576: About ... entity of Type?
Post by George Bruseker (14 December 2021)
Dear all,
Recently work is on-going on a new property 'represents thing of type' which is distinct from 'represents' (again that particular vs categorical distinction).
https://cidoc-crm.org/Issue/ID-476-pxx-represents-entity-of-type
I am confronted with cases of an information object being about not a particular thing but a category of thing... in my case event types but I guess it could be object types. Of course the existing 'about' property is sufficient but it doesn't allow to differentiate that it is not just a type but about an as yet unknown X which was of type Y (about a particular thing that is known categorically'; eg .Sales Record about 'Sale Event') ... It seems to me similar to the other new property we are working on already.
Does anybody else have cases like this? Any interest in a new parallel property like that OR a solution that requires no new properties but also doesn't require semantic back flips to understand?
Best,
George
Post by Martin Doerr (1 January 2022)
Dear George,
I think this is a very good idea. There are thousands of archaeological publications listing items etc. of certain types, often with reference to museums keeping them, but library practice will only register overall aboutness. Museum records cite such publications explicitly, but the inverse has never been exploited systematically.
Cheers,
martin
Post by George Bruseker (6 January 2022)
Dear Martin,
I'm glad to hear you have encountered such cases as well and find it a potentially good path to explore. I find myself in a conundrum thinking about (my own) proposal because of the vagueness of the word 'entity'.
What I find typical in many modelling exercises regarding bibliography is that aboutness can be about the following real world, particular things:
subject - E55
geographic location / place - E53
person / corporation - E39
So already there is typically some breakdown of aboutness into different main kinds of real, particular things that a work can be about.
When I reflect on the property 'represents entity of type', the examples are all of endurants. This may be an accident but it makes me wonder about whether a new specialization of aboutness to refer to 'things of type' would best follow this general pattern (the entity could be a perdurant, endurant, place etc.) or if the property would better be more specialized. (is about temporal event of type, is about persistent item of type, is about place of type).
It just occurs to me that in the context of retrieval if the property were not more specific to perdurants / endurants etc. then it could be quite difficult to sort out if you want to find works about events of type vs works about things of type vs works about places of type etc. This is not a problem when the aboutness property is about a particular because we can use the class to differentiate. 'Given me E73 about E39' automatically culls the data down to the aboutness regarding actors. If the new property pointed to E55, then we would not have this facility.
So... hopefully still a good idea but seems to have some complications to be thought through.
Best,
George
Post by Martin Doerr (6 January 2022)
Dear George,
We use the "type" because it implies necessarily if it is a perdurant or endurant, person etc. If it does not, it is ill-defined, and has no place in a Thesaurus (see the AAT). If the categories of a thesaurus fit the CRM is a mapping problem.
No problem for retrieval at all. Just a programmers job.
Place types are relatively rare, such as "river" "lake", "city". The UMLS system e.g., listed some decade ago I think 10 or 20 million types, but less than 100 properties, corresponding to at most 200 classes. Therefore, P2 is not "cheap", but it does never replace meaningful properties.
Best,
Martin
Post by George Bruseker (7 January 2022)
Hi Martin,
We use the "type" because it implies necessarily if it is a perdurant or endurant, person etc. If it does not, it is ill-defined, and has no place in a Thesaurus (see the AAT).
Unfortunately, I find many thesuri don't follow good ontological principles, so we are forced for workarounds, but okay.
If the categories of a thesaurus fit the CRM is a mapping problem.
No problem for retrieval at all. Just a programmers job.
That job can be very expensive!
Place types are relatively rare, such as "river" "lake", "city". The UMLS system e.g., listed some decade ago I think 10 or 20 million types, but less than 100 properties, corresponding to at most 200 classes. Therefore, P2 is not "cheap", but it does never replace meaningful properties.
"Cheap" in "Cheap and Cheerful" is not a denigration... https://www.macmillandictionary.com/dictionary/british/cheap-and-cheerful
Cheers,
George