Wikidata:Property proposal/subclass of with uncertain existance
subclass of with uncertain existance edit
Originally proposed at Wikidata:Property proposal/Generic
Description | equivalent of this class where it's uncertain whether the instances of the class actually exist; can also be used if it's certain that they don't exist |
---|---|
Data type | Item |
Example 1 | human whose existence is disputed (Q21070568) → human (Q5) |
Example 2 | organisation that may or may not be fictional (Q113584205) → organization (Q43229) |
Example 3 | object of uncertain existence (Q115471146) → castle (Q23413) |
See also | fictional or mythical analog of (P1074) |
Motivation edit
This property will be a subproperty of subclass of (P279). We don't want that human whose existence is disputed (Q21070568) appear in queries for human (Q5) and as such it's problematic to actually let it subclass human (Q5). At the same time, users regularly try to add the subclass at normal rank. I propose this property so that this relationship can be well modeled. ChristianKl ❪✉❫ 11:40, 27 November 2022 (UTC)
Discussion edit
This discussion comes out of https://www.wikidata.org/wiki/Wikidata:Project_chat#buildings_whose_existence_is_disputed
I'd want as few classes generated as possible, so I'd push the class as high as possible, so not object of uncertain existence (Q115471146) but a class of architectural structure (Q811979) called "proposed architectural structure", and when its used for a castle have a instance of (P31) qualifier with value castle (Q23413). I'd use it for historical references to buildings, not mythical ones, but the dividing line is blurred here. For that reason I'd use 'proposed' rather than 'fictional', as its thought to be part of the historical record, not an invention as part of a work of fiction. "Proposed" might not be the best word, as that could include building plans, perhaps "suggested" would be better. Vicarage (talk) 12:05, 27 November 2022 (UTC)
- Regardless of the case of castle we still have organisation that may or may not be fictional (Q113584205) and human whose existence is disputed (Q21070568). ChristianKl ❪✉❫ 16:09, 27 November 2022 (UTC)
Comment The problem is that the class of all X which might exist is not a subclass of the class of all X which do exist. Depending on how you understand “might exist” (i.e. does this only include X with unproven existence or all?), it’s the other way round at best. So “subclass of” is a poor naming, in my opinion, and making it a subproperty of subclass of (P279) would encourage sloppy modelling. That said, I do Support the idea of creating a property for this kind of linking, but I’d recommend naming it differently – “counterpart with uncertain existance to” or “uncertain existance version of” perhaps. --2A02:8108:50BF:C694:55AC:C874:B5ED:83CD 11:12, 28 November 2022 (UTC)
- Oppose This can be perfectly well modeled with subclass of (P279) by creating superclasses that include the possibly fictional - something in between human (Q5) and person (Q215627). ArthurPSmith (talk) 18:33, 28 November 2022 (UTC)
- Of course it can – the question is whether requiring non-fictional classes to link to the fictional ones is proportionate (given that there is a potential fictional counterpart class for everything) and whether having to change the non-fictional class when a fictional class has to be introduced is considered appropriate. --2A02:8108:50BF:C694:343B:A504:8B48:52D9 21:41, 29 November 2022 (UTC)
- Comment For fictional and mythical entities we have fictional or mythical analog of (P1074). Not sure if this could be broadened to include entities that might exist (I'm a rather hesitant myself, right now). But if the line to mythical entities might be blurred, it could be the best to do so. Notified participants of WikiProject Fictional universes - Valentina.Anitnelav (talk) 11:13, 30 November 2022 (UTC)
- If fictional or mythical analog of (P1074) should be broadened one could relabel it something like "has analog in real world" or "has analog in reality" (inspired by the current Russian label of fictional or mythical analog of (P1074)) - Valentina.Anitnelav (talk) 11:39, 30 November 2022 (UTC)
- Comment How would the proposed property handle transitivity? Would all subclasses of a class "with uncertain existence" inherit the uncertain existence property? — The Erinaceous One 🦔 07:09, 1 December 2022 (UTC)
- That’s exactly the question with “might exist”. If “might exist” means that existence is (has to be) uncertain, then every subclass would inherit the uncertain existence, so the “not uncertain” class could not be a subclass. If we understand “might exist” as “may or may not exist”, then the “not uncertain” subclass simply comprises those items where the “may not” part does not apply (and its subclasses, in turn, would comprise some of those items). I don’t see a problem with transitivity in either case, but the former wouldn’t be useful for the problem at hand. --2A02:8108:50BF:C694:254A:DF6C:D37E:B93F 12:57, 1 December 2022 (UTC)
- @2A02:8108:50BF:C694:254A:DF6C:D37E:B93F: if you want to vote and comment and have your votes respected it would be great if you would register an account instead of just commenting anonymously. ChristianKl ❪✉❫ 13:34, 1 December 2022 (UTC)
- At the moment I’m fine with commenting. But I do consider creating an account. --2A02:8108:50BF:C694:254A:DF6C:D37E:B93F 17:17, 1 December 2022 (UTC)
- @2A02:8108:50BF:C694:254A:DF6C:D37E:B93F: if you want to vote and comment and have your votes respected it would be great if you would register an account instead of just commenting anonymously. ChristianKl ❪✉❫ 13:34, 1 December 2022 (UTC)
- One solution would be to let human (Q5) and fictional human (Q15632617) subclass human whose existence is disputed (Q21070568) as Arthur proposes. I created a thread in the ontology Wikiproject for that idea. If we decide for that this property wouldn't inherit anything, otherwise we likely still want to explicitely say for every subclass what their equivalent is and not use value hierarchy property (P6609) to specify transitivity. ChristianKl ❪✉❫ 13:26, 1 December 2022 (UTC)
- Would we have to have a parallel class tree for each class? In other words, would we also make "dogs who may be fictional", "castles that may be fictional", "airplanes that may be fictional", etc.? This approach seems overly burdensome. — The Erinaceous One 🦔 01:41, 2 December 2022 (UTC)
- I would expect that we would create items like "dogs who may be fictional", "castles that may be fictional", "airplanes that may be fictional" whenever someone actually needs them. At the same time, it likely would produce an extra tree and that might be bad for SPARQL query performance.
- Creating this property would do the job without needing an extra tree. ChristianKl ❪✉❫ 12:06, 3 December 2022 (UTC)
- Would we have to have a parallel class tree for each class? In other words, would we also make "dogs who may be fictional", "castles that may be fictional", "airplanes that may be fictional", etc.? This approach seems overly burdensome. — The Erinaceous One 🦔 01:41, 2 December 2022 (UTC)
- That’s exactly the question with “might exist”. If “might exist” means that existence is (has to be) uncertain, then every subclass would inherit the uncertain existence, so the “not uncertain” class could not be a subclass. If we understand “might exist” as “may or may not exist”, then the “not uncertain” subclass simply comprises those items where the “may not” part does not apply (and its subclasses, in turn, would comprise some of those items). I don’t see a problem with transitivity in either case, but the former wouldn’t be useful for the problem at hand. --2A02:8108:50BF:C694:254A:DF6C:D37E:B93F 12:57, 1 December 2022 (UTC)
Comment WD astronomers have already wrestled with uncertainly and came up with https://www.wikidata.org/wiki/Special:WhatLinksHere/Q72053253 and create a small number of classes with caveat names. This doesn't seem scalable. I still think have a single new class 'uncertain status' with aliases 'unconfirmed as', 'suggested site of' 'maybe fictional', 'maybe mythical' etc and then using the normal classes as objects would be best. So a frog might be 'uncertain status' 'frog' and 'prince'. Invisible to most queries, easy to select systematically for completest ones Vicarage (talk) 05:10, 2 December 2022 (UTC)
- How do you plan to make sure that it is invisible to most queries? Should "frog" have normal rank and "uncertain status" (or "entity of uncertain existence") a preferred rank? Or should "frog" have a deprecated rank and "entity of uncertain existence" a normal one? I'm not sure about this as it seems to utilize ranks for something they were not designed for. But your comment reminds me of a property proposal some time ago: Wikidata:Property_proposal/Fictional_instance_of_real_object. One could phrase it something like "described as instance of" and use it for any kind of thing - hypothetical persons, mythical entities, hoaxes, fictional entities, etc. This would also be an alternative to your proposed use of instance of (P31) as a qualifier. But while this would be just an addition to many projects this would mean a complete overhaul of the model that is now (successfully and quite consistently) used for many years for fictional entities. This would be quite disruptive for the fictional universe Project and come up with a completely different model. I just wanted to throw this into the room as another possibility. - Valentina.Anitnelav (talk) 11:05, 2 December 2022 (UTC)
- For a concrete example Cae Tump Placename 2, Gladestry (Q38611878) has instance of (P31) of object of uncertain existence (Q115471146) with qualifier instance of (P31) castle (Q23413)
- Try it!
SELECT DISTINCT ?item ?itemLabel WHERE { *: SERVICE wikibase:label { bd:serviceParam wikibase:language "[AUTO_LANGUAGE]". } *: { *: SELECT DISTINCT ?item ?possible WHERE { *: ?item p:P131 ?county. *: ?county (ps:P131/(wdt:P279*)) wd:Q5566244. *: {?item p:P31 ?statement2. *: ?statement2 (ps:P31/(wdt:P279*)) wd:Q23413. *: } *: UNION *: {?item p:P31/pq:P31 wd:Q23413.} *: } *: } *:
- Finds Cae Tump Placename 1, Gladestry (Q39011444) as a castle, and Cae Tump Placename 2, Gladestry (Q38611878) as a possible castle. This would require a instance of (P31) change to accept a new class of possible things. Vicarage (talk) 12:12, 2 December 2022 (UTC)
- If someone runs a query for "How many humans lived past 110?", that query should automatically avoid counting people in the bible with mythical status. Generally, if people do a SPARQL query they don't want objects that might not exist to be included by default. Requiring everyone to runs a query to explicitely think about not showing objects that don't exist creates an unnecessary burden. It also makes the SPARQL query more complex and thus has performance costs. ChristianKl ❪✉❫ 12:11, 3 December 2022 (UTC)
- This would be solved by “human” being a subclass of “human who might (not) exist”, and “biblical figure” being a subclass of the latter, too. If someone wants to include humans who might not exist, they query for (transitive) instances of “human who might (not) exist” instead of “human”. The problem with parallel class trees persists, though. --2A02:8108:50BF:C694:545:E0DA:374F:594A 20:14, 8 May 2023 (UTC)
- Not done, no consensus of proposed property at this time based on the above discussion. Regards, ZI Jony (Talk) 05:16, 25 January 2024 (UTC)