Tomastvivlaren

🔒📴

Residential heating and how to describe the Internet

Unless I'm mistaken, we have met before, right? If so, do you still live where the landlord excels in heating up your entrance hallway?

Since I'm new here (Wikidata), and this page is kind of public, I'm sticking to English for the benefit of other readers. But what actually led me here was your recent addition of a determination method (P459) statement on the Internet (Q75), as that seems to contradict dogmatic practice here. Maybe you can shed some light on your intentions with that edit, and give me some feedback on other issues as well? --SM5POR (talk) 18:11, 27 May 2020 (UTC)

Sorry, I confused you with Joshua (one of them), who wrote about the hot landlord. But I did encounter both of you in the same community, I think. And my feedback invitation remains. --SM5POR (talk) 18:46, 27 May 2020 (UTC)

@SM5POR: Thnks for nice messages. I do not remember any landlord troubles. Occasionally I am in Uppsala (which Google associates with your name) but I live 300 km up north. I don't think you need feedback on your edits, but perhaps I need. I wrote most of the Swedish version of sv:Template:Databox and sv:Module:databox, which I hope other languages would adopt. Someone suggested me to never show {{P|P2283)} in the infoboxes, since for the most part it is rubbish, not suitable in encyclopedic articles. I followed the advice. Then I replaced Internet -> is using -> IP with Internet -> follows the standard -> IP. Not for dogmatic reasons, but just to somehow show IP in the sv:Internet article infobox.Tomastvivlaren (talk) 19:17, 27 May 2020 (UTC)

Ok, double confusion on my part it seems, since I was thinking about thomastvivlaren.se which I now understand hasn't been your website. Anyway, while I believe I can manage without feedback on my edits, I'd like to avoid spending time correcting minor errors by others and instead cooperate to make this great 80+ Mitem heap of Internet fertilizer in our Creative Commons front yard even greater. I don't have any bots working for me, but I want to get involved where my efforts would be most beneficial (look at my user page if you want to know what I'm planning). So far I have mostly been entertained by bots, and they don't exactly come up with a new idea every day. --SM5POR (talk) 19:46, 27 May 2020 (UTC)

I tend not to believe in broad generalizations (such as the claim that 98.43% of all statistical data is simply pulled out of thin air), and before even reading your earlier discussion about it, I wanted to find out myself. You can have a look at my results and experiment with the queries if you like. The most frequent class of items using the uses (P2283) property is a figure skating competition (Q2990963), and municipality of the Czech Republic (Q5153359) sort of stands out among the others, while dog (Q144) didn't even make it among the top 200 classes. I agree with both of you though that uses (P2283) is probably best kept out of generix infoboxes, given its apparently loose definition, and your solution with complies with (P5009) is probably better than "uses" when the subject is some technical infrastructure (In Soviet Russia, the Internet uses you)...

But the determination method (P459) edit is what caught my eye, as the statement doesn't even make sense, like saying the Internet simply "is" according to some partcular determination method. It's a property qualifier to be used with other properties only, such as the detection of exoplanets by means of Fourier analysis (Q1365258) or having your apartment stylished according to feng shui (Q170537) principles. --SM5POR (talk) 07:32, 28 May 2020 (UTC)

Of course you are right. I removed it.Tomastvivlaren (talk) 16:43, 28 May 2020 (UTC)

Proto-Norse and Runic Swedish

(Due to deteriorating eyesight, I just enjoyed misreading the Spanish description: "protolengua de las lenguas nördicas"...)

I see, thanks for pointing the difference out to me. I was simply trying to satisfy the inverse statement on Runic Swedish (Q10657846) that was flagged as incomplete: follows (P155) Proto-Norse (Q1671294). But then I suppose that one should be removed instead? And what about followed by (P156) Early Old Swedish (Q10546113), or said to be the same as (P460) Old Danish (Q12330003)? Maybe you can come up with a better Swedish description than "forntida germanskt språk talat i Sverige", as that one risks leading other editors astray? I see that you have indicated Runic Swedish (Q10657846) being part of (P361) Old Swedish (Q2417210), which is probably better, but there are now two statements saying the same thing; isn't that redundant?

I also replaced the Gregorian date types for start time (P580) and end time (P582) with their Julian counterparts, but as I have no sources, I cannot verify the dates. Would you say "8. century" is an appropriate replacement for the year 800, or should it rather be the 9th century (considering our knowledge of Runic Swedish)? And what about the implied high precision in "1225"; what is the state of research there? If you want to indicate some other uncertainty range than a full power of ten number of years, that may be possible using ± format (such as 1225±25), but I haven't checked.

Speaking of nördic languages, what do you think about Gothic Wikipedia? I mean, wouldn't even a Klingon Wikipedia make better sense, given the number of current speakers? Nothing beats citing a GNU Free Documentation License (Q22169) clause in original Klingon... --SM5POR (talk) 06:22, 8 July 2020 (UTC)

Hint for English-language readers: "nördic" is pronounced "nerdic". --SM5POR (talk) 06:35, 8 July 2020 (UTC)

As usual, reality is way ahead of my imagination. Apparently there was a Klingon Wikipedia 16 years ago, but it's now gone (at least at that address). Now, I happen to think that a Sumerian or Egyptian Hieroglyphic Wikipedia would be the best thing since sliced bread, at least for its educational value; is somebody working on that yet? Consider the hilarious ramifications of the entertainment industry taking legal action against websites dedicated to swapping modern pop song lyrics (inspired by Jukka Ammondt (Q762263), of course) or television series subtitles translated into old Sumerian. The kids would learn the language in no time, just to tell someone 𒆠𒆪. --SM5POR (talk) 07:19, 8 July 2020 (UTC)

The Klingon Wikipedia was moved and renamed The Klingon Encyclopedia (Sov qawHaq tlhab). Ok, I think we can leave it there... --SM5POR (talk) 07:49, 8 July 2020 (UTC)

Subclass, part of or place? The Databox module

A somewhat related question @SM5POR: Do you think that a language should be "subclass of" (p279) a language family, or be "part of" (p361) a language family? Why? I think I have seen both, allthough the first is much more common, but many wikipedians prefer the latter. They dislike wikidata terminology such as subclass of, instance of, etc, in infoboxes.

If you are interested in the reasons for my recent editing of language related objects, you may read Databox multilevel lists (databox looks better if each object only has one predecessor, one follower, etc), history of Swedish language (provides sources that support my edits), debate about Databox in language articles. and debate about Databox in Krim gothic. Tomastvivlaren (talk) 11:35, 8 July 2020 (UTC)

I'm right now working on a longer piece about the relationship between Wikidata and Wikipedia infoboxes, and it would be interesting for me to see some practical examples of the latter. Looking at sv:Mall:Databox#Vad_mallen_visar_och_inte_visar, I wonder where those fundamental rules are implemented. Can you give me a pointer to the actual source for things like excluding human (Q5) from the unlabelled instance subtitle? Is it a global definition, or local to Swedish Wikipedia? If local, is there some meta-documentation for implementing project-specific infobox templates? --SM5POR (talk) 01:52, 10 July 2020 (UTC)

The Swedish version of the Databox module now hides "Human", because the historicity of some semi-legendary figures (for example saga-kings) is a grey zone, and because "Human" is self-evident in articles about historical people. You never would write "Human" in a non-wikidata based biography article. I am also considering to hide the word "Event" for similar reasons. Hopefully the Swedish extended version will be spread to the other countries that use Databox (especially Africaans and French).Tomastvivlaren (talk) 09:16, 10 July 2020 (UTC)

Ouch!!! 😉 No, I don't mean to ridicule; I can tell there is plenty of hard work and skilled thinking behind this code... Yet I think there is room for structural improvement. Or rather, some structural improvement should lead to more room for thinking and less code to maintain in total, by a factor of... 300, perhaps? Ok, probably not that much, as Swedish indeed is a major language on Wikipedia, right up there with English and... Cebuano? Never mind. What I mean is, over 30 KB of imperative programming for one language seems a bit unmanageable to me. The English counterpart is about half that size, yet I think that is too much as well. But I do agree about the functionality being desirable; the only language where it would make sense to label named persons as "human" would be one primarily spoken by robots. Or other species. Then simply hide the Q5 class label by default, and let R2D2 re-enable it for Robot Wikipedia only (if he/she/it wants to). In that way, you won't have to consider Afrikaans, French, Gothic, Cebuano or any of the other 300 languages specifically, assuming they rely on the same language-independent Infobox/Databox module.

Do you know if it would be possible for us to build a private template module for trying out some new functionality, as a proof-of-concept? Would it require some administrative privileges to make it run even on our User namespace pages only? Does the module have to reside in the Module namespace? The kind of changes I have in mind can simply not be achieved by gradually adjusting the existing infoboxes, which is just as well, since you don't want to touch them until you have tested the new mechanism and verified it will scale also performance-wise. And in case it doesn't work, nobody needs to bother about it. --SM5POR (talk) 16:37, 10 July 2020 (UTC)

At the Swedish Databox template talk page people have suggested improvements to Databox that I have tried to implement. They complained about terms like "instance of" and "human", so I removed them. Many also complain about "sub-class of", they say Databox impose Wikidata terminology on Wikipedia, but I have been unwilling to remove subclass of (P279), and I have not found any better replacement for that word than "subtype" (Swedish undertyp).

You are quite right, the lack of cooperation between language versions on developing Lua code means a huge amount of unessecary work. For example, some Swedish Lue modules have Swedish variable names etc, limiting its use in other languages. The Wikipedia core is written in PHP and developed centrally, only by a restricted group of people, while Lua Modules and Wikicode templates are developed locally by anyone. Yes, you can develop a private version, for example in sandboxes such as [1], [2] and [3]. The module code can be automatically synchronized between language versions, but that feature is seldom utilized. I think it was a mistake to not store Lua modules centrally. A new project called "Abstract Wikipedia" is supposed to include common storage of lua code, to my undestanding.

The Swedish version of Databox can potentially be spread to other languages, and that is my desire. But I need someone to cooperate with that is willing to work on this. I have placed a sandboxed copy ot it at https://www.wikidata.org/wiki/Module:Databox/Sandbox for testing purposes. The English documentation page template:Databox/doc now shows examples both from current code and the sandboxed Swedish version, allowing people to compare the result. First of all, the Swedish documentation should be translated to English. Some local adoptions of the code to other languages may improve it, but are not necessary to make it work.Tomastvivlaren (talk) 14:03, 11 July 2020 (UTC)

Ok, is that (I'm now referring to the apparently widespread dislike of Wikidata property labels) why the ontology is such a mess? I don't mean the linguistics ontology specifically, but in Wikidata in general. Fortunately it's not a complete disaster, but even a mere 5 percent messy linking tends to wear off on the rest, yielding an inconsistent overall impression regardless. Thanks for pointing this contributing factor out to me; I wasn't aware of it. Then...

Houston, we have a problem!

I'll get back to your specific question about the language properties shortly, after we have sorted this out (you may be well aware of this already, but now I'm aiming at a slightly wider potential audience, so please excuse my verbosity):

The database and the property system have been designed from principles based on Class logic (Q1297858) and ontology (Q44325) (the latter theory having been developed for IKT systems as ontology (Q324254)).

The properties thus describe fundamental relations between objects of any kind, physical entities (space, time, energy, matter) as well as abstract concepts (number, meaning, language, science). They enable programmers to write code that automatically retrieves information to produce factual statements about the Universe as we know it. It may eventually allow AI systems to seek out information independently of any human operator and draw conclusions from it.

Thus the statement wd:Q3957 wdt:P31 wd:Q48907157 means that "a town is a specific instance of a section of populated place" (an element in a set), wd:Q3957 wdt:P279 wd:Q702492 means that "a town is a kind of urban area" (a subset), and wd:Q3957 wdt:P361 wd:Q6256 that "a town is part of a country" (a constituent component).

These are inherently different properties which normally can't replace each other; it would be wrong to classify a town as "a smaller kind of country". Lawyer and plumber are both instances of profession; you can hire a lawyer or a plumber, but you cannot hire a profession. Likewise with subclass and part; a lawyer's left arm isn't a more specialized or less competent lawyer, but a part of the lawyer.

The properties and items have been given natural language labels and descriptions as a convenience primarily to database query programmers, not Wikipedia end-users or anybody else (and we may thank Grace Hopper (Q11641) for her clever insight that programmers shouldn't talk to computers in numberspeak only).

A program doesn't care about the property numbers being cryptic or inaccurately labelled; it depends only on the numbers being correct and consistently applied to appropriate objects. For all the program cares, the P31 property may be labelled "instance of", "part of", "green penguin at large" or "better a bottle in front of me than a frontal lobotomy"; the program will work fine just the the same.

But replace P31 with P279, and the program will go bonkers, regardless of how P279 is labelled.

The differece between subclass of (P279) and part of (P361) isn't one of linguistic nuance between "subclass" and "part"; those are two different relations with different semantical properties. Which one to use must be determined by their inherent meaning, not by what looks best in Wikipedia infoboxes.

For that reason, I consider it a bad idea to use the property labels even as default infobox labels. When you design an infobox for Wikipedia usage, you should be required to spell out explicitely which Wikidata properties to use for which infobox contents, and how to label them.

This is why I'm thinking of designing a different kind of infobox, where you have to list those properties you are prepared to present to the user (and specify what labels you want in each language), rather than have any property values appear automatically as soon as they are added to the Wikidata item. Wikidata is designed to accept an arbitrary amount of data about each item; all of it probably shouldn't be listed in the Wikipedia infoboxes without editorial oversight and discretion. --SM5POR (talk) 18:46, 11 July 2020 (UTC)

I'm now returning to the issue of how to treat languages:

If, in a linguistics infobox, you want to state that Swedish is an East Scandinavian language, you could for instance say something like ?lang wdt:P31 ?fam; "Language:" ?lang; "Language family:" ?fam (as you can probably tell, I don't have the slightest idea of what an infobox definition looks like, but I hope you get the general idea of what I'm trying to say).

The above example assumes that in the linguistics domain, Wikidata editors have agreed to treat language families of any size or age as classes, and languages as instances of those classes. When you consider what is part of (P361) a language, I would assume it's used for things like grammar, vocabulary, pronounciation, writing system, orthography; i.e. concepts that are required parts of any language, but may vary arbitrarily between different languages.

I actually find it a bit unclear how part of (P361) is supposed to be used for abstract concepts. It's pretty easy for physical objects; a bicycle for instance has parts like wheels, frame, pedals, saddle and so on. Take a few of those parts away, and you will have a broken bicycle.

Compare this to an abstract domain like linguistics; could you use part of (P361) to describe the Nordic (or North Germanic) languages as "part of" the Germanic branch of the Indo-European family of languages? Sure, if you agree that when one of those "parts" disappears, like the East Germanic languages (Gothic, Old Preussian or whatever they were identified as), the Germanic branch is no longer complete, but forever "broken". So... No (I think you agree with me here). A language isn't a "part" of a language family any more than a bicycle is a part of a mode of transport (you can still walk or take the bus if your bicycle is stolen).

But if a language were hypothetically to lose, say, its vocabulary, or its grammar, it would be a broken language, no longer usable. All parts aren't necessarily essential; you could have a language without a writing system, and it could still be used orally, but it wouldn't be as useful as one that also has a written representation (same with a bicycle without a working light).

If you can divide a group of items (a class) into an arbitrary number of smaller groups containing some of those same items, those smaller groups are properly called subclasses, not parts. You may have any number of coexistent subclass hierarchies, such as centum and satem languages, historic and modern languages, European and African languages, written and non-written languages, official and suppressed languages, and they are all classes, as long as they group the same kind of items: Individual languages. Which class hierarchies to actually use is something you should discuss and agree on within some appropriate Wikidata WikiProject.

Then you have dialects... As there is no natural distinction between "dialect" and "language", linguists will simply have to agree which is which according to some criterion, such as "a language is a dialect with an army" or whatever. Once you have that definition, you need to map their mutual relationship to Wikidata properties, and I would suggest considering each dialect an instance of (P31) a language.

However, as that requires the language to be defined as a class of something, you need to identify its parent class, which I'd say is not its superior class of languages, but the class of dialects, an entirely different class hierarchy, but still within the linguistics domain. Thus Swedish will be an instance of (P31) the East Scandinavian languages, and at the same time a subclass of (P279) the Scandinavian dialects, and perhaps a part of (P361) of the Swedish elementary school curriculum (the latter along with English, Mathematics, and physical training, to name a few).

The group of dialects spoken in Norrland would constitute a subclass of (P279) of Swedish (unlike individual dialects, which are instances). And then the sociolects, slang, cocodode lolanongoguagogesos, baby talk, what have you... --SM5POR (talk) 18:54, 11 July 2020 (UTC)

┌────────────────────────────────────────────────────────────────────────────────────────────────────┘ Thankyou for pedagocical explanations.

To be concrete: Do you think that the Wikidata relations shown in following lists are correct, or do they include the 5% of mess in the ontology that you mention? It would be very helpful for me if you would find time to help me with a quality control of the content that Databox presents, before we discuss how it is presented.

• List of Databoxes at Swedish Wikipedia that shows "Sub-class of".
• List of Databoxes at Swedish Wikipedia that shows "Part of".
• List of Databoxes at Swedish Wikipedia that shows "Instance of".

I have noticed that "Part of" sometimes is used to show that places are part of larger places (for example, all Swedish civil parishes such as Bureå socken are "part of" provinces such as Västerbotten), and some villages such as Hjoggböle are part of parishes such as Bureå socken. Is that correct usage? Other villages such as Gissjö have "Location" in a civil parish, and buildings may have "location" in residential areas. Places may belong to current "administrative units" such as Swedish municipalities and counties. See

• List of Databoxes at Swedish Wikipedia that shows "Place".

Regarding how to present the information: Can you give a concrete example of a possibly better row title in the infobox than these three property names? For example, in the case of Herculaneum, I think it is better to write "[[comune of Italy|Current comune]]: Ercolano" than to write "Administrative unit: Ercolano". In the case of Stockholm, I think it is better to write "[[Counties of Sweden|County]]: Stockholm County" than to write "Administrative unit: Stockholm County". But this would require huge amount of manual work to adopt the row headers. Wikidata does not give much help in automatically identifing that Ercolano is a comune, and Stockholm County is a county.

I have also been considering to adopt the Databox code, to allow users to replace property names with other row titles, or to hide certain properties, but I need good examples when this helps. It is not diffucult, but my impression is that it would solve very few problems. Problems at Wikidata should be solved there, but such a feature might make it tempting to overrule Wikidata locally. And then the whole purpose of using Wikidata for infoboxes is lost.

Wikidata is a shared project. Both by people like you who wants to build a knowledge base that future AI can use to automatically generate rules and one day take over the world. ;) (Am I correct?) And by Wikipedians who only wants to automize the translation of infoboxes in their articles. Without the Wikipedians, Wikidata would not grow and be the worlds largest ontology. We have to somehow agree between these two cultures, and make both groups happy. Property names and infobox row headers may now always have the same name, but they must correspond. To avoid confusion, to default use property names as row headers, and only occasionally replace them, is a very good idea.

Many Wikipedians say they do not understand the things you try to explain here, and are unwilling to learn wikidata. But they are unqilling to contrbite to wikidata, so any mess in the ontology can not be their fault. And it can not be because of generic infoboxes like the "Databox" (only used in French, Afracaans and Swedish wp), "Univerecal Infobox" (essentially only used in russian wp) or the "Wikidata Infobox" (mostly used at Wikimedia Commons). These are used in very few articles - only a few houndred or thousand articles at each of these Wikipedias. Many wikipedians are still looking at generic infoboxes with skepcism because the code has been way too simple, and still the first two of these three templates can not show citations.

No, I think the bot owners are to be blaimed. The major reason for the mess is that bots have imported a lot of incorrect data from non-wikidata-based infoboxes at Wikipedia, sometimes from poor article versions at small language versions of Wikipedia, without requirirng reliable external sources, and without even including the external sources if they were provided. Wikidata and its bots are sloppy in requiring sources. Few relationas are imported from other external ontologies. Bots do not regularly import changing data from external sources, for example every year, meaning that the data often is out of date. Tomastvivlaren (talk) 23:21, 11 July 2020 (UTC)

Short name

You have added a short name (P1813) for geographic coordinate system (Q22664) (in Swedish). However, as I understand this property, it's meant for official names of persons, places, organizations, standards and so on (such as FDR, JFK, EU, Soviet Union, IPv4, MIDI, and perhaps kWh), but not for shorter expressions of natural language terms in general (we may have borderline definition issues with things like GUI and TV).

I have left it in place though in case you had some good reason to add it, maybe a school project (are you a teacher)?

For most language-specific properties like scientific terminology, alternate words and so on, I think they belong in the labels/aliases section of the item, as well as in the growing Wikidata:Lexicographical data effort to systematically describe the vocabularies of the world's languages. However, I'm not too familiar with the latter, as evidenced by me just failing to add a semantical link from Lexeme:L242100 to coordinate (Q3250736); I must read a bit more about it to find out how it's supposed to be done.

Speaking of coordinates, and your earlier mention of a place some 300 km north of me, maybe you will appreciate this? --SM5POR (talk) 09:50, 8 July 2020 (UTC)

Hi. Regarding Q22664, it was an experiment that I forgot to revert. Thanks for reminding me. However, I have added many Swedish shortnames to properties, for example "koordinater" to coordinate location (P625), because that will force the Swedish version of template:Databox to show the short name instead of the full property label "geografiska koordinater". "Koordinater" is the usual parameter name (or row header) in Swedish infobox templates. Especially the Swedish property labels are often unnecessary long, but people have asked me to not try to shorten the labels.

I did not understand if your query shows the centroid of a place? Tomastvivlaren (talk) 11:10, 8 July 2020 (UTC)

@SM5POR: You've made quite advanced queries. Inspired by you, I made this simple query: https://w.wiki/WT8 . It shows a list properties that have Swedish short names, in most of the cases created by me, aiming at the Databox template.Tomastvivlaren (talk) 23:37, 8 July 2020 (UTC)

I wrote a comment to address a number of points above, but as I were to post it, I ran into an editing conflict with you on my previous thread about languages, and decided to put my follow-up on hold (I have saved it locally on my laptop) until we have resolved the broader issue here.

Well, I'm still learning the query language, and quite often I get a timeout running my queries, so many of them are left unfinished while I try other approaches. As you may tell from my en:User:SM5POR page, Wikidata is a project well suited to my long-term aims, but I'm probably lucky not having discovered it until now, as I may approach retirement within a few years... Before the World-Wide Web, I had an idea of what an Internet Web would look like. Only one thing -- it wasn't at all like the World-Wide Web. It was more like Wikidata... --SM5POR (talk) 02:58, 10 July 2020 (UTC)

In my youth, I had some ideas about database user interfaces with bidirectional links. Perhaps this will be accomplished by the new wikidata API (see https://www.mediawiki.org/wiki/Wikidata_Bridge), which will make wikidata more integrated in WIkipedia. Tomastvivlaren (talk) 09:33, 10 July 2020 (UTC)

Wikipedia vs Wikidata

The thread above has grown a bit too large for me to engage in, so I'm detaching the issues that I think haven't yet been resolved...

Ok, maybe I shouldn't blame the "mess" (which isn't objectively verified) on Wikipedia editors, but I should take aim at the robots instead. Of course all kinds of users should coexist on Wikidata, and Wikipedia is indeed a great source of information to get Wikidata started. But besides the 300 language editions of Wikipedia, there are the other Wikimedia Foundation projects (Commons, WIkisource etc), and Wikidata is growing on its own as well. If I understand the statistics I have seen correctly, quite a large share of the items are entries for scientific articles that I don't think have passed via Wikipedia, but they have been directly added to Wikidata by robots.

All public information isn't suitable for the encyclopedic presentation format, but may still be indirectly useful in various database searches. If you are able to present pretty much everything Wikidata knows about an item in an infobox now, it's only because most of that information comes from Wikipedia in the first place. By showing any new property values by default, you may effectively limit the amount of information editors dare to add in order for Wikipedia not to explode. If you have to actively select those property values to show, then you specify what the infobox should look like first, and request the property values you need for that later, and you will never have to deal with properties falling over you unasked for.

I have a few ideas of how to go about this, wherefore I want to get started with sv:Modul:Sandlådan/SM5POR/Unibox and sv:Användare:SM5POR/sandlåda, but I don't quite understand how they interact. Maybe I need to clear some cache or what? --SM5POR (talk) 21:07, 12 July 2020 (UTC)

Aha! When everything else fails, read the fine manual... Now I got the invocation correct! We are in business. --SM5POR (talk) 14:46, 13 July 2020 (UTC)

Consistent ontologies

I might find it a bit easier to walk through you property assignments for a quality control if all the items were duly linked, and there were some kind of checkboxes. My main problem though is that I don't know all those subjects by heart; I would have to read a lot of articles to get the picture.

By the looks of it, I'm afraid there are quite a few errors. The two most essential properties are instance of (P31) and subclass of (P279); together they form the backbone of Wikidata, and I believe that every single item (excluding redirects) should ideally be indirectly reachable by traversing successive numbers of those two properties only. An item with neither an instance of (P31) nor a subclass of (P279) property is an orphan, regardless of any other properties it may have (very much like how a human orphan may well be a part of (P361) a community without having either an employer or parents to "belong" to).

While subclass of (P279) is a transitive property, instance of (P31) is not. If the East Scandinavian languages are Germanic languages, and the Germanic languages are Indo-European languages, then the East Scandinavian languages are Indo-European languages too (they are subclasses of each other). On the other hand, if Donald Trump is a human, and human is a species, then it does not follow that Donald Trump is a species, because being an instance of (P31) is a non-transitive relation!

Therefore you may typically have any number of subclass of (P279) links in a property path, but it's important to keep track of the number and individual placement of any instance of (P31) links, as each one of them represents a jump up or down an abstraction level.

As a rule of thumb, anything that has a proper name written with a capital initial also in the middle of a sentence (in Swedish) is an instance only, not a class (there may be exceptions, but I can't come up with one right now). It must however be an instance of (P31) a class, not of another instance-only item (they constitute the leafs of the ontological class tree). Anything that does not have a proper name is probably a class, but may simultaneously also be an instance of something.

Therefore, any named individual (singular), be it a person, a particular thing in your possession (even if unnamed), an organization, a place, a planet, a galaxy far far away, a universe, or an elementary particle that I met yesterday who presented itelf as "Bogon Googoolplex^42", is an instance of (P31) of the corresponding class of humans, things, organizations, places, planets, galaxies, universes, or bogons (maybe it was a mistake ruling that class items should be labelled in singular; writing their plural forms make them stand out much better).

Then there are the tricky cases, such as those written with a capital initial in English but not in Swedish (say, languages). I'll continue with thoae later. --SM5POR (talk) 06:39, 13 July 2020 (UTC)

For a quick first check of the queried lists you referred me to, look through the second column of the subclass of (P279) and instance of (P31) lists, which shows the parent class, and apply the plurality test: Can you easily turn each of those terms into their plural forms with no essential change in meaning, or are they already written in plural? This is where you have to help me; is Old Norse ("fornnordiska") a single language, or a family/group of languages? Maybe a group of dialects? Did anyone in the past ever speak "Old Norse", or did they speak one of several "Old Norse languages"? The latter is a requirement if it's supposed to have either subclasses or instances of itself.

Then look at the fourth column of the subclass of (P279) list, can those items be written in plural as well? I'd put a question mark on any linguistic item that doesn't end with "språk" (hinting at plural, languages). Is Proto-Indo-European and old language, or a family of proto-languages? Likewise with Quicksort, Järnvägsalgoritmen (the definite form is a giveaway), Isoniazid, sionism, kolera, Uniform Resource Locator, PH, Balkanhalvön, Cree; some are more obvious errors than others, but this is not an exhaustive list anyway.

I'm not asking you to fix those errors found immediately, as doing so now may temporarily introduce other inconsistencies or disrupt your Infobox algorithms. You should rather write down an ideal model for how those classes of items and their corresponding infoboxes should relate to each other, and see if you can adapt your code to gracefully handle a period of transition. For that reason already, I think you will need to move away from "If item I has property P1 with value V, show value V but call it (some nice thing), show other properties and their values as-is" to "If databox D wants to say Nice Thing, look up property P1 or, if that fails, use property P2 as a fallback, else drop Nice Thing as not present".

I suppose this may make your code more complicated, not less, and it may even turn out to be too much work. But if you help me get started with my Unibox experiment, maybe I can learn enough to come up with an easier approach later. --SM5POR (talk) 09:45, 13 July 2020 (UTC)

When I look closer at the instance of (P31) list, it actually looks mostly ok. One general objection though: (The) history of pestilence, architecture, Denmark, Spain and so on are hardly classes, and neither is history in general (again, try the plurality test; it would be bizarre to talk about "Danish histories" when referring to historical events in Denmark).

Instead, those events would be proper to call part of (P361) the history of Denmark, and as part of (P361) is also a transitive relation, it's quite ok (recommended, even) to make historical events part of (P361) the appropriate historical period, and make each period part of (P361) history in general; likewise with the history of each nation, field of human activity, illness, technical invention, or the Earth (with events such as the Cambrian explosion (Q32919) that could properly be considered part of (P361) Cambrian (Q79064), though I see now Wikidata has a problem with the number of leap days to consider when calculating "dates" some 500 million years ago).

But to keep the ontologists happy, each event also has to be an instance of (P31) of something, and you could as a last resort make them instances of occurrence (Q1190554), but please don't; it's already crowded up there. And creating a parallel class structure "event during period so-and-so" would seem pretty redundant, so I would rather suggest arranging them according to an entirely different qualitative criterion, such as natural events, political events, violent events, religious events, extraterrestrial contact events, ELE... --SM5POR (talk) 14:15, 13 July 2020 (UTC)

Part vs Quality

I have come to the conclusion that my analogy between languages (communication devices) and bicycles (transportation devices) doesn't work very well, in particular with respect to the part of (P361) property. One characteristic of being physically part of a bigger entity is that you still exist as a separate unit within that larger whole, and your own constituent subparts typically don't interact with subparts elsewhere in the parent object. There are exceptions to this, such as municipalities, communities and individuals of different countries sharing boundaries or dealing with each other across national borders, like the golf course straddling the Swedish-Finnish border in Haparanda-Tornio, or Malmö (part of Sweden) and Copenhagen (part of Denmark) sharing a very physical bridge with all the technical infrastructure involved.

This is not the case with orthography, writing system, pronunciation, vocabulary, or grammar, which I previously suggested could be considered part of (P361) a language. The distinction is that the orthography interacts with pretty much every minor aspect of all the other properties, such as what letters or other graphemes your writing system has, how the voiced dental sounds or foreign loan words should be spelled, use of space to delimit compound word parts, and so on, likewise in between the other properties. You cannot isolate the grammar and "remove" it from the language, thereby leaving a language with a "hole" in it, but it would no longer be a language at all.

Therefore I think has characteristic (P1552) is a much better generic property to use for associating a language with each of those mutually entangled features I mentioned, unless there is a more precise property defined in Wikidata. There are a few properties defined specifically for the lexicographic section (lexemes, senses and forms), but they are probably limited to that section and should not be used with items in general.

I still find part of (P361) appropriate to use for a language being part of a school curriculum, or a letter being part of an alphabet, as you can still identify individual parts and move them around (like how W is now considered a letter of the Swedish alphabet, which it wasn't before 2006). --SM5POR (talk) 12:41, 15 July 2020 (UTC)

Swedish vs English

Here is a side issue which caught my attention yesterday, when trying to find out which leap day calendar would be appropriate for the Cambrian (Q79064) period:

There are two classes of days, called non-holiday (Q819073) and day of the week (Q41825) respectively. I was first confused about which one I should use, as non-holiday (Q819073) was labelled "veckodag" in Swedish, and day of the week (Q41825) had an alias "vardag". It turned out that all (or at least a major part of) the other languages had it labelled and described the other way around. I corrected the obvious error, added a description and some aliases in Swedish, after which I went on to investigate the individual days of the week. I was of course not satisfied with what I saw, wherefore I went through the list to change instance of (P31), subclass of (P279), and part of (P361) properties according to my liking, such as adding appropriate qualifiers like Monday being a "weekday" except on holidays (such as Christmas). You may want to have a look to see whether you agree with my statements.

Then today, I checked my edit history to see whether anybody had made additional changes to these items, perhaps even reverted or overwritten my edits. There were no such changes, and I took another look at Sunday (Q132) to see whether there was something else I could improve,

Then I discovered what Sunday was supposedly named after (P138) in various languages, such as Spanish or French. It said Mister (Q177053) ("Mister", "Herr").

What? Even though I do know some Spanish, I didn't immediately make the connection: In those languages Sunday is typically associated with God (Q190), who in at least Swedish is sometimes titled "Herren". But I didn't know this linguistic connotation was so widespread... To see Sunday described as being named after "mister" just looks bizarre to me. And it has been like that since 14 July 2013 and nobody has bothered to fix it. If God had been considered a woman, would we now have a discussion of whether to use Miss, Mrs, Ms, or some other form of polite address..? --SM5POR (talk) 19:07, 15 July 2020 (UTC)

Audio files needed

This polite plea for some audio files thoroughly made my day. I cannot possibly imagine a more deserving request. I wonder what it might sound like in Gothic? --SM5POR (talk) 07:02, 18 July 2020 (UTC)

Description of "scalar"

I'm skeptical about this edit (and the ones following it) to scalar (Q1289248):

[4]

Ultimately, these edits replaced "elements of a field, e.g. real numbers, in the context of linear algebra" (which had a double space, makes sense mathematically, but doesn't appear to me to necessarily be the best description around) with "single element field, e.g. real number, in the context of linear algebra".

It's not clear to me what that's supposed to mean. A field (Q190109), in mathematics, always has at least two distinct elements. And an everyday "field" is a coordinate (Q3250736). That leaves field (Q185674), which is usually said to have values, not elements.

However, I think it makes sense to leave scalar (Q1289248) to mathematicians, and use scalar quantity (Q181175) for the physics concept. Streetmathematician (talk) 14:49, 31 October 2021 (UTC)