Wikidata:Property proposal/banned in

From Wikidata
Jump to navigation Jump to search

banned in/from

[edit]

Originally proposed at Wikidata:Property proposal/Organization

   Not done
Descriptionexcluded / prohibited -- illegal / illicit / forbidden -- [in (place), at (time), by (agent)]
Represents!permitted in /better representation of one element of the odd property dissolved, abolished or demolished date (P576) (which has type constraints currently making it impossible to add to e.g. slavery (Q8463))
Data typeItem
Template parameternone
Domainitem
Allowed valuespre-existing item
Example 1chewing gum (Q130878) → Singapore (Q334) (qual: point in time)
Example 2Daily Mail (Q210534) → English Wikipedia (Q328) (qual: point in time)
Example 3alcoholic beverage (Q154) → mosque (Q32815)
Example 4sandal (Q131704) / clog (Q167633) / shoe (Q22676) → Hindu temple (Q842402)
Example 5blasphemy (Q200481) → Alsace-Moselle (Q2840162) (qual = until / end 22/12/2015)
Example 6slavery (Q8463) → Qatar (Q846) (qual = date = April 10, 1952 = 16th Rajab 1371)
Example 7serfdom (Q103350) → Austrian Partition (Q129794) (qual = date = 16 April 1848)
Example 8benfluorex (Q421695) → France (Q142) (qual = date = November 2009) (reference)
Example 9English Wikipedia (Q328) → Turkey (Q43) (qual = inception + end)
Example 10Ulysses (Q6511) → United States of America (Q30) endtime = 1933
Example 11Ulysses (Q6511) → United Kingdom (Q145) endtime = 1936
Planned usevery generally useful

Motivation

[edit]

I'm not sure this is exactly in the right place, as this is a general property of diverse items.

It seems strange that we cannot say that chewing gum is/was banned from Singapore, clogs from temples, alcohol from mosques, the Daily Mail, the Sun, Breitbart (and soon RT) from en.wp. Of course some of these should probably have time qualifiers (targeted immigration bans, COVID travel bans...) This is a more willful version of "excluded from" which I also didn't see in the pre-existing properties. Now it might make more sense to exclude the preposition but then we would need to have "in" / "from" & "by"... which gets trickier. Moreover it's not clear to me that saying that chewing gum was banned by Goh Chok Tong would be overly interesting. Hope I've filled this out correctly. SashiRolls (talk) 23:33, 9 May 2020 (UTC)[reply]

FWIW: in terms of adding landing sites for further enrichment (other semantic roles), a bot could be written to create a page Ban of Qx from Qy item for every instance of Qx (has property banned from Qy). (The opposite might also be true, but first the property needs to be created.) SashiRolls (talk) 21:52, 18 May 2020 (UTC)[reply]

Discussion

[edit]
  •  Comment this doesn't feel quite like the right way to model this, though I agree some approach to recording such relationships would be useful. Maybe there should be an item for each of these ("ban of x from y") which captures all the relevant data about the proscription? ArthurPSmith (talk) 19:13, 11 May 2020 (UTC)[reply]
  • My intuition is with ArthurPSmith. Apart from that it might be misleading to use banned to describe the relationship of EnWiki and Daily Mail. The judgement of that RfC is a complex one that says that the Daily Mail can't be used as reliable source which is not a blanket ban. ChristianKl ❪✉❫ 11:29, 12 May 2020 (UTC)[reply]
    Yes, the problem of en.wp's "deprecations" raises interesting questions. Reliable sources talk about "bans", while RS/N (as I insisted on the Daily Mail TP some months ago) talks about a "deprecation". In practice this is playing on words, because in general once outlets like the Sun and Breitbart have been banned / "deprecated" a removal campaign is begun (and the outlets are sometimes even added to the black list as Breitbart was). In other words, don't be fooled by the clever lingo: it is a ban, though exceptions can be argued for.
    In terms of ontology, the idea of changing a relation like Qx has p(banned in Qy) to an event like ban (cf. block (Q2125156)) has (p of chewing gum (Q130878)) has (p from Singapore (Q334)) strikes me as wrong-headed confusing because for the information to be noted on the original data"card" (here chewing gum (Q130878))) would involve all sorts of pirouettes. The proposal is much more straightforward as written IMO. (otherwise we'd be creating convoluted/overwrought/unresolved NPs (noun expansion) as stand-alone cards for events, if I've understood correctly. For information, this suggestion was created after looking into WikiLambda... whose goal seems to be to encode such relators (prepositions and verbs are basically relators).
    I do agree banned from is probably better than banned in (but that is cosmetic). Similarly "banned" could be created at the same time as "banned from" if an "event" reading rather than a "has property" reading is wanted, but as I said above, I'm not sure that Goh Chok Tong banning chewing gum is overly interesting. (bans would be better for the reciprocal relation, where it exists...
    Alternately I see that there is a property: does not have property, which could be combined with "permitted in", though that latter property (permitted in) also does not yet exist. SashiRolls (talk) 18:35, 12 May 2020 (UTC)[reply]
    If you have a law/RfC that bans something, I think it would be better to have an item for the law/Rfc and then has a statement "bans X" and one for the jurisdiction in which the law applies. ChristianKl ❪✉❫ 09:23, 15 May 2020 (UTC)[reply]
    Why? SashiRolls (talk) 12:51, 15 May 2020 (UTC)[reply]
    @SashiRolls: Wikidata only allows limited data attached to a statement - in particular qualifiers cannot have their own qualifiers. By "reifying" a statement of an event/action etc. into its own item you can have a much richer description, which is what seems needed here. ArthurPSmith (talk) 18:13, 15 May 2020 (UTC)[reply]
    I'm still stumped. In the examples above only one qualifier was used, though I can see that if you wanted to specify when something was banned and who banned it you would need double qualification (or two properties, banned in place & banned by agent)
    benfluorex (Q421695) has the property (banned in France (Q142)... qualifier: O° as medicine 1° since November, 2009). (ref-url)
    You suggest making an item Q99999999 (Ban of medical use of benfluorex in France). Interesting. What further steps would then be necessary to connect that new item to the pre-existing data item benfluorex (Q421695), which is the main subject of Q999999999, and which is the data item that has the property banned in France (Q142) that it would be good to know about.. I'm just curious how your proposal would end up being written onto the data item needing further description... ? What query could then be written to find all medications/chemical compounds banned in France in 2009? I appreciate the time you are taking to understand my question. This is perhaps a more pressing example than those above since Wikidata is telling people that benfluorex (Q421695) is a medication at the moment without mentioning it doesn't seem to be approved anywhere?-- SashiRolls (talk) 20:22, 15 May 2020 (UTC)[reply]
    @SashiRolls: So I think a somewhat analogous case would be seasons of sports leagues - for example 2006–07 Fußball-Bundesliga (Q27896). That is an intermediate item between the sports team and the league for a particular year. You can easily query for the linkages - for example all seasons connecting teams and leagues here. Is this what you're asking for? ArthurPSmith (talk) 02:08, 17 May 2020 (UTC)[reply]
    basic semantic triple model
    @ArthurPSmith:: Why don't you do what I asked rather than providing what you think might be "a somewhat analogous case"?
    (season is not a predicate of sports team) <-- actually, it probably is... but one which requires its own table in a traditional db. (sashi / 19 May 00:08ish)
    (banned in France is a predicate of benfluorex)...
    I have created the requested item so that you can show me how it is less kludgy than a direct statement of the property on the appropriate item (benfluorex): Ban of Mediator in France (Q94603373) I've provided all the references you need there for the item. Could you provide a SPARQL query now showing all the medications banned in France in 2009? What further steps did you need to take in order to make that happen? (At the moment nothing indicates (on the ban page) that Médiator is a medicine, for example... You get that "for free" by adding banned in Q142 as a property to benfluorex (Q421695))
    -- SashiRolls (talk) 02:20, 17 May 2020 (UTC)[reply]
    No I'm not suggesting making an item called "Ban of medical use of benfluorex in France". There's likely a regulatory act that produced the ban. A regulatory act that has an inherent name. The regulatory act also has a bunch of interesting meta-data.
Blasphemy might be banned in 1000s of locations. Adding 1000s of statements to a single item is bad for Wikidata. There also might be 1000s of things that are banned in a given location. ChristianKl ❪✉❫ 21:42, 19 May 2020 (UTC)[reply]
Yes, just think of dry counties... I'm sorry that the RW Wikidata tries to represent is messy. :) SashiRolls (talk) 05:27, 20 May 2020 (UTC)[reply]
  •  Comment Structurally this is not that dissimilar to educated at (P69) (past participle + preposition). While we could create Education of Qx at Qy items (by bot) which could then be enriched with all sorts of ancillary info (thesis adviser, subjects studied, degree obtained, matriculation date, praise in Latin, etc.), that wasn't done for the predicate "educated at". Why follow a different tack here? SashiRolls (talk) 22:04, 18 May 2020 (UTC)[reply]
  • It's structurally very dissimilar. I know of no person who's educated at (P69) at 100 universities. On the other hand if you just count the countries that have banned slavery you can easily go over a hundred. At the same token you can also find hundred (likely even more) things that are banned in a given country. ChristianKl ❪✉❫ 21:52, 19 May 2020 (UTC)[reply]
  • Hm... My claim was that it was "not that structurally dissimilar": (past participle + preposition), not that the content (or number of members) was the same. At the moment, can you provide a list of the dates slavery was banned in all states using Wikidata? If so, how? If not, why not? I notice slavery (Q8463) currently has 43 identifiers... I look forward to reading your improved proposal for representing the data on this page. You might want to look at Thirteenth Amendment to the United States Constitution (Q175613) with its signatories too in terms of your concerns about multiple statements. That item mentions "abolition" or "ban" only in the label. Its main subject "slavery in the United States", fails to capture the (effect / main subject) of the 13th amendment... How, @ChristianKl:, would you improve on this unsatisfactory representation? Keep in mind that this is just one case of the predicative relationship (property = banned (in Qy)). (Here is a more transient example (similar to "dry counties" above because of the embedded predicative relationship (sale of)). Here is another.) SashiRolls (talk) 04:53, 20 May 2020 (UTC)[reply]
  • When it comes to database modeling structure a N:1 relationship is structurally dissimilar to a 1:N relationship. If we would ignore database engineering concerns and only look at the ontological relationship when creating Wikidata properties, that would hugely damage Wikidata.
The 13th amendment could have a statement with P:newProperty:bans slavery (Q8463) applies to jurisdiction (P1001) United States of America (Q30). ChristianKl ❪✉❫ 13:51, 20 May 2020 (UTC)[reply]
Cool, we seem to be making progress. 1 person is educated at N schools (secondary school, undergrad, masters, phd, law school, med school, etc.), and 1 THEME is banned in 1-N places (by 0-N texts). Reworking it for the verb (ban) rather than the past participle (banned), as I think you seem to prefer, we also have [0-N texts ban(s) 1-N themes in 1-N places] (e.g. city ordinance + federal law ban smoking in federal buildings). Am I following you? (I'm not sure why you mentioned the standard db structure test). Of course, we often won't have the name of a text banning many THEMEs. The example of Benfluorex is an example... the Lancet article I cited above doesn't mention the name of any regulatory text. For the story of alcohol being banned in mosques, or sandals being banned in Hindu temples, I'm not sure there is a specific hadith or Hindu sutra that does the banning (there may be, or the ban on alcohol may be derived from the general proscription of alcohol as haram, and the ban on sandals may be derived from the general cultural practice that you don't wear shoes inside in Asia). Thoughts? Also do you have another example of a property that is an active form of a verb as your proposed "bans" is above? SashiRolls (talk) 19:11, 20 May 2020 (UTC)[reply]
Whether something is an active form of a verb is irrelevant (how the English language models the relationship is not a reason). For the purposes of Wikidata database performance 5 is 1 and not N in terms of database complexity. ChristianKl ❪✉❫ 19:50, 20 May 2020 (UTC)[reply]
First, I don't understand what you mean when you say 5 is 1 and not N. Could you explain what you mean by that, please? SashiRolls (talk) 20:24, 20 May 2020 (UTC)[reply]
Second, it seems to me a very distinct semantic role is added by the active voice.
If we model the predicative relation on the active voice (bans) we need to have an item for the agent (or the text produced by agents with standing to ban).
If we model the predicative relation on the participle/adjective + prep (banned in), you don't need to have an item for the agent (Qz), just for the patient/theme Qx (what is banned) and the place (Qy) where it is banned.
Of course nothing prevents us from creating both an active property (agent, patient, place, date) and a passive property (patient, place, (date)). It's true that this is clearer (that both could be created) than in the case of educated / educated at. (Did Zellig Harris (alone) educate Chomsky at MIT? Did University of Edinburgh (alone) educate JK Rowling?, ...)
Finally there is the problem of Qx banned from Qn (activity), that could dangerously end up on DLP. (e.g. player banned from (playing in) league due to testing positive for anabolic steroids) SashiRolls (talk) 20:39, 20 May 2020 (UTC)[reply]
It's alright if you don't understand anything about database modeling, but then how defering to what people who do understand on it?
The cost of an edit is proportional to the size of the edited item. As far as your 'second' goes, Wikidata is currently at capacity for the amount of edits that it can do. Your proposal (for educate) might result in halving the amount of edits that can be done per day on Wikidata. ChristianKl ❪✉❫ 08:17, 21 May 2020 (UTC)[reply]
I did not make any proposal about "educate". I pointed out that it was not logical to create it. As for the rest, see your talk page, please.SashiRolls (talk) 09:02, 21 May 2020 (UTC)[reply]
While waiting for the property creators to wave their wiki-wands, I've created this (beginning of a ) workaround based on the suggestions above. I've added this to a couple other entries (Emancipation reform of 1861 (Q1192212) & Slave Trade Act (Q7539168)). Now what shall we do for example 6 above (Qatar)? Create an item for the obscurely-named FO 371/98464? I assume this is a UK Foreign Office file, since it is owned by the National Archive. It is not the original text of the proclamation, but a translation... SashiRolls (talk) 13:38, 21 May 2020 (UTC)[reply]
Nobody suggested using instance of (P31) here and adding it just because there's no property that you like isn't how to go about it. ChristianKl ❪✉❫ 12:08, 22 May 2020 (UTC)[reply]
I'm surprised disappointed to see an administrator removing a solution that works. I imagine you will explain your reason for removing a working solution and either provide a better one or restore the working one... SashiRolls (talk) 15:28, 22 May 2020 (UTC)[reply]┌────────────────────────────────────────────────────────────────────────────────────────────────────┘

See also: https://w.wiki/RaN SashiRolls (talk) 18:08, 22 May 2020 (UTC)[reply]

We have a policy for property proposals that centers around new solutions to model content being decided by property proposals. Having a community decision for a given solution means that outside parties can rely on that solution existing in the future and that works within the framework of Wikidata without causing problems. As admin it's part of my role to prevent people from trying to circumvent policies. ChristianKl ❪✉❫ 09:59, 26 May 2020 (UTC)[reply]
@ChristianKl:: Please stop assuming bad faith. From the point of view of someone trying to find simple ways of extending Wikidata's descriptive power this really feels like an admin lording their power without doing the intellectual modeling work... Here, I asked you a question about the banned book (Q65770793) item made by another person in July of 2019 (shortly before its dominating class banned work (Q66475652) was created by a different person. Both of these items have since been edited by the same administrator in September of last year, which leads me to believe nobody has found anything intrinsically wrong with them. I'm afraid you didn't look carefully enough at the example to see the real problem (I'm really not sure what could have confused you regarding end time (P582).) Try looking again. Also, as I said on your talk page, you may find some ideas for your concerns about your "authorization of subject" property proposal in the ODRL 2.1 prohibited / permitted modules for DRM I added for you to look at. SashiRolls (talk) 18:35, 26 May 2020 (UTC)[reply]
  •  Comment I have created a list of all the items (banned books, works, objects, etc.), classes, and properties I've turned up so far in this effort to improve WikiData's power to describe bans. There is also a bit of feedback on the experience farther up on the same page. Any property creators willing to put on their magic hats, yet? What still needs discussion? Active voice versus passive voice, still? SashiRolls (talk) 21:56, 23 May 2020 (UTC)[reply]
  • I'm not confident this is the right way to model this and the tone in this discussion suggests we aren't going to get to a productive conclusion here. There are subtleties this proposal ignores (is the ban of slavery the same kind of ban as of sandals? are sandals banned in mosques or is it the wearing of sandals?). I would prefer we link the banning to the "act" that does the banning but I do see that that wouldn't always be easy. The arguments about wikidata DB load to me seem hollow without some kind of reference to a developer statement. BrokenSegue (talk) 17:35, 27 May 2020 (UTC)[reply]
  • @BrokenSegue: It would be nice if we would have perfomance implication well documented so that it's easy to link them. Just because they are not well documented however doesn't change that they are important for the functioning of Wikidata. ChristianKl ❪✉❫ 15:31, 11 June 2020 (UTC)[reply]
I was very pleased that the above proposal for an active voice "prohibits" was created in November 2020 using my example of the 13th amendment above. I see that we still have a proliferation of nominalisations of a basic predicative relationship:
If we are meant to pursue this practice of adding a nominalisation (banned practice, banned person, banned medication, banned religion, etc.), and then using the "instance of" property to find these items, I suppose that could work. My question: is there an easy way to automate retrieval of the *.wp pages of, say, "censored books" (166 members) and add "instance of banned book" to those entities, or do these things need to be done manually (to add qualifiers like dates, jurisdictions, etc.)
I assume this is a better solution than creating two pages for the laws that banned and then unbanned, say, (parts of) Les Fleurs du Mal in France, or multiple page for the town counsels/high schools which have banned Slaughterhouse Five?
I see too that the property "prohibits" wasn't linked to the entity "slavery" on the wikidata page for the 13th amendment. I still think it is a mistake to prioritize the active voice over the passive voice... I do agree that we can drop "in" from "banned in" since the qualifier "applies to jurisdiction" covers "in". However, the passive voice is as, if not more, useful than the active voice. That has not changed. SashiRolls (talk) 21:18, 9 December 2020 (UTC)[reply]
Correct me if I'm wrong but based on the example given in "prohibits" we'll need to search for items that have the property "prohibits" linked to the entity slavery in order to pull the start date out of the law banning slavery in order to make a list of the dates when slavery was abolished and the places where it was abolished? SashiRolls (talk) 21:34, 9 December 2020 (UTC)[reply]
No, "instance of" is not meant to be used that way. If there are two laws, then yes that means item that legalizes it again can use prohibits (P8739). ChristianKl ❪✉❫ 21:45, 9 December 2020 (UTC)[reply]
Please explain. I would have thought Joyce's Ulysses was an instance of a banned book. SashiRolls (talk) 21:56, 9 December 2020 (UTC)[reply]
It mixes attributes of how the book relates to other entites into instance of (P31) while there's no suggestion that this should happen on https://www.wikidata.org/wiki/Wikidata:WikiProject_Books ChristianKl ❪✉❫ 22:08, 9 December 2020 (UTC)[reply]
Thank you for making my suggested edit to the 13th Amendment page ( § ). Your mention of a "mixture" is due to the fact that "banned" is a predicative relationship, I think. Now that you've deleted the collaborative work on Ulysses ( § ), how do you propose to encode the "knowledge equity" in *.wp about banned books that is just "ripe for the encoding" at the various *.wp? 166 at en.wp for example. SashiRolls (talk) 22:24, 9 December 2020 (UTC)[reply]
I'm not sure why you use the term "knowledge equity" in this context. From the point of knowledge equity, information being "ripe for encoding" means that it's not valuable to add the knowledge because the information can already assessed easily. "Knowledge equity" is about giving that information which is not easily accessed by people space. ChristianKl ❪✉❫ 00:07, 10 December 2020 (UTC)[reply]
Note that as of this writing Slaughterhouse-Five is not listed on any of the various *.wp Wikidata should include. Le Quart Livre & Les Fleurs du Mal are only listed on fr.wp, etc., etc. There have been a lot more than 166 banned books in the course of history, as I'm sure you know. Banned / Prohibited books are also just one part of this proposal. There is at least one "instance of a banned compound" mentioned in the examples... SashiRolls (talk) 00:32, 10 December 2020 (UTC)[reply]
There's a tradeoff between quantity and quality. Requiring this syntax helps with the quality as the statements that we do have about bannings will include what did the banning.
"instance of a banned compound" seems like a statement that's unclear. Are certain people banned from entering it? Is the existance of it banned (and thus it should be demolished)? When we require the origin of the ban to be stated it's easy to access that information.
Batch imports from Wikipedia have the additional problem that they are essentially unsourced for many contexts which makes it less desireable to optimize our syntax to make that easy. ChristianKl ❪✉❫ 13:21, 10 December 2020 (UTC)[reply]

┌────────────────────────────────────────────────────────────────────────────────────────────────────┘

I thank you for expressing your opinion, Christian. Perhaps others will agree with you, but I do not. Yesterday, you overruled the consensus of two people who worked on the entity Ulysses without any discussion ( § ). In trying to understand why, I imagined that you might be thinking that whether a work is banned or not is an accidental rather than an essential characteristic.
Research shows that this does not seem to be how Wikidata works. Take for example the banned book Slaughterhouse-Five (Q265954). Not only are there lists of awards received on the entry page, but there are also lists of awards it was nominated for (note the passive voice of both P166 & P1411). These are accidental or a posteriori type statements, not essential or a priori type statements.
I will ask you again: how would you encode bannings in the same way consensus has chosen to encode awards that books have won or been nominated for? Or if you don't want to answer that question... why do you think there should be two different encoding methods for awards and for bannings? What is gained by being inconsistent?
Another (minor) conceptual limitation with "prohibits" (incidentally) is that it is a present tense. A text that has been abrogated -- such as the text banning six of the elements of Les Fleurs du Mal -- no longer "prohibits" anything. This problem does not exist with the past participle "banned", since it is abstracted out of time... (is/was) banned. Perhaps you are beginning to see why I believe the passive voice is superior to the active voice (it focuses on the effects on the entity rather than on an external agent).
I think most people would (perhaps naively?) agree it would be absurd to change P166 "awards received" to gave award and P1411 "nominated for" to nominates for. I would ask those reading this discussion if they think this is an oversimplification, or if is is an accurate analogy to the position Christian is taking. SashiRolls (talk) 19:22, 10 December 2020 (UTC)[reply]


 Not done Over a year with no consensus. BrokenSegue (talk)

Retrieved from "https://www.wikidata.org/w/index.php?title=Wikidata:Property_proposal/banned_in&oldid=1523789621"