Property talk:P6609

Documentation

value hierarchy property
property which specifies less precise items than the indicated value, for which statements using the subject property would still be true

Data type

Property

Example

instance of (P31) → subclass of (P279)
headquarters location (P159) → located in the administrative territorial entity (P131)
occupation (P106) → subclass of (P279)
afflicts (P689) → anatomical location (P927)

Use to indicate transitivity?[edit]

(@Pintoch:) Certain properties, such as located in the administrative territorial entity (P131) and subclass of (P279) have had this property applied pointing to themselves, indicating the same thing as instance of (P31) transitive Wikidata property (Q18647515). Is this an appropriate use of the property? Should all transitive properties use this form? --Yair rand (talk) 21:22, 24 March 2019 (UTC)[reply]

@Yair rand: I would say that this is appropriate indeed, unless we estimate it is an unwelcome redundancy. It's not like we have that many transitive properties anyway, so it does not incur a huge duplication cost. − Pintoch (talk) 08:49, 25 March 2019 (UTC)[reply]

There are currently 14 instances of transitive Wikidata property (Q18647515). --Yair rand (talk) 21:11, 25 March 2019 (UTC)[reply]

Naming ?[edit]

« Transitive over » seems way better to me as a name. It’s more common, « property for value hierarchy » seems completely cryptic and ad hoc. Anyone (dis)agree ? WikiProject Properties has more than 50 participants and couldn't be pinged. Please post on the WikiProject's talk page instead. WikiProject Ontology has more than 50 participants and couldn't be pinged. Please post on the WikiProject's talk page instead. author TomT0m / talk page 13:31, 22 May 2022 (UTC)[reply]

Hello TomT0m

, Tu as fais une traduction du libellé en français vers l'anglais. Dans les faits, « transitive over » est déjà un alias en anglais. De plus, quand on utilise cette propriété — dans d'autres propriétés, comme indiqué par allowed-entity-types constraint (Q52004125) — je préfère qu'il soit indiqué propriété clairement. Cela évite son utilisation abusive. Voir aussi la section précédente avec transitive Wikidata property (Q18647515) pour ne pas confondre. Cordialement —Eihel (talk) 14:05, 22 May 2022 (UTC)[reply]

@Eihel ben « propriété pour la hiérarchie de valeurs » moi je trouve pas ça clair du tout. Déjà il y a des trucs qui sont transitifs mais certainement pas hiérarchique, les équivalences par exemple … Une relation comme « les amis de mes amis sont mes amis » ce serait transitif, mais ça n’établit pas de hiérarchie, et c’est transitif. Après si je rajoute « ce qui est à moi est à mes amis », par exemple, ça établit une relation de « transitivité de la possession par l’amitié », mais je vois pas du tout en quoi ce serait décrit par « propriété pour la hiérarchie de valeurs » author TomT0m / talk page 14:13, 22 May 2022 (UTC)[reply]

@TomT0m, @Eihel: I agree that "value hierarchy property" is a non-intuitive name, but I disagree about "transitive over" even being correct. What value hierarchy property (P6609) describes is not which properties (or rather values) are transitive, but which values are inherited from other items, regardless of inheritance mechanism. It just so happens that the only mechanism referenced by value hierarchy property (P6609) so far is via an arbitrarily long path of a repeated transitive property relation.

The distinction between transitivity and inheritance was pointed out by @Yair rand at Help talk:Basic membership properties#Subclass transitivity over three years ago. I returned to that thread the other day to see if anybody had taken notice yet, and only now discovered the link to the property proposal discussion which I hadn't noticed before. Lo and behold; there was this inheritance property which I have been searching for since I joined Wikidata in 2020, and the reason I didn't find it was that I kept searching for words like "inherit" and similar.

I would therefore suggest renaming or aliasing value hierarchy property (P6609) to something that includes this word, such as "value inherited via", "property inherited from", "inheritance path" etc. That also opens the door for adding other paths than transitive ones where values are inherited, such as the single instance of (P31). How would an application otherwise determine that "wdt:P31/wdt:P279*" is a source of implied property values, but "wdt:P31/wdt:P131*" is not? Are you aware of any applications using value hierarchy property (P6609) today? How do they work?

I have no opinion on the appropriate property label in French (but I may help suggesting one in Swedish, when we agree on what it should mean). --SM5POR (talk) 11:00, 7 December 2022 (UTC)[reply]

@SM5POR « transitive over » is different from just « transitive » !

Sorry I failed to have included a reference but the ontotext ontology use a transitive over property whose definition matches, the example of « located in » and « subregion of » is given. author TomT0m / talk page 11:14, 7 December 2022 (UTC)[reply]

I can accept "transitive over" being different from "transitive" (and thus retract my argument that it's incorrect), but how can you then indicate that an item inherits a number of property values from its "wdt:P31" class, but not from its "wdt:P31/wdt:P31" meta-class? Like queen Victoria being a human, and human being a species that has evolved over millions of years, but queen Victoria not being a species that has evolved for that long? Is the transfer of property values from a class only to all immediate instances of that class (but not further, since instance of (P31) is not a transitive relation) somehow automatic, and will not need to be specified on a per-property basis?

I can hire a lawyer, and lawyer is a profession, but I cannot hire a profession... SM5POR (talk) 12:11, 7 December 2022 (UTC)[reply]

@SM5POR Can’t see your problem, subclass of (P279) is transitive, instance of (P31) is not, but instance of (P31) is transitive over subclass of (P279). instance of (P31) is however, neither transitive nor transitive over instance of (P31). So there is no risk to import properties from the metaclass to the instance. author TomT0m / talk page 13:35, 7 December 2022 (UTC)[reply]

subclass of (P279) transitives, is the same as the math property of « less than » : if 1 is less than 2 and 2 is less than 3, then 1 is less than 3. This is no surprise of course. With "subclass" it amounts to "if all dogs are animals, and animals are beings, then dogs are beings" (dog subclass of animal and animal subclass of being, then dog is a subclass of being)
instance of (P31) transitive over subclass of (P279) is something different, of course : if Rex is a dog, and if all dogs are animals, then Rex is an animal.

None of these rules, 1 or 2, applies to metaclasses. If "Rex" is a "dog", and "dog" is a "being class", then … nothing. So there is no risk to inherit whatever, by these rules at least. (edit author TomT0m / talk page 14:07, 7 December 2022 (UTC))[reply]

I may have misinterpreted value hierarchy property (P6609), assuming genre (P136)value hierarchy property (P6609)subclass of (P279) alone would be sufficient for Agenre (P136)G to be inherited by every indirect subclass of A. If that's not the case, but you also need subclass of (P279)value hierarchy property (P6609)subclass of (P279) to make subclass of (P279) transitive, then it should work as I had expected.

But none of this happens automatically; any application that depends on implied (inherited) property values will have to look up value hierarchy property (P6609) for every property wanted, and to apply this algorithm uniformly (say, in Wikipedia), some supporting library routines are needed. right? Is there such a library already, or is it under development somewhere?

Because with values being obtained simultaneously via multiple transitive properties, you will also have to determine how to combine different values found via different paths, say by adding them together, replacing more distant values with closer ones, or whatever makes most sense for a particular property and value class (such as languages spoken or time zone applied in a geographic area; I'd add the languages but replace the time zones).

Then we have property paths affected by qualifiers... Thank you for the explanations anyway! SM5POR (talk) 15:53, 7 December 2022 (UTC)[reply]

@SM5POR Wikidata does not implement any of these rules indeed, even the transitivity of "subclass of" which is pretty standard, for example in RDF/RDFS. We have to deal with this

manually, for example, with property paths in SPARQL. It’s a problem but I think community have never really been able to have a concensus on how to get to something like that to ask for the devteam, we still struggle to implement something as reverse properties …

> I may have misinterpreted propriété pour la hiérarchie de valeurs (P6609), assuming genre artistique (P136)propriété pour la hiérarchie de valeurs (P6609)sous-classe de (P279) alone would be sufficient for Agenre artistique (P136)G to be inherited by every indirect subclass of A

No it does not say anything about the subclasses of A at all. It says something about the relationship between A and the superclasses of G.

> Because with values being obtained simultaneously via multiple transitive properties, you will also have to determine how to combine different values found via different paths, say by adding them together, replacing more distant values with closer ones, or whatever makes most sense for a particular property and value class (such as languages spoken or time zone applied in a geographic area; I'd add the languages but replace the time zones).

That depends of the usecase. Typically, for example, a usecase of these rules could be to ask « does lung cancer affects the respiratory system », but we only know it affects « lung ». But the lung is a part of the respiratory system, so the « affects » property should probably be transitive over « part of » for. If such cases no problem about whatever value to « chose » (for what, display in an infobox?)

There is no such lua library as far as I can know, there is stuff to deal with transitive properties in the french fr:Module:Wikidata, especially the special case of instance of (P31) and subclass of (P279), but I think that’s all to my knowledge. There is also Module:PropertyPath to iterate in lua over the possible values, that can help implement some rules. I also created a module to deal with union of (P2737)

/ disjoint union of (P2738)

that is used in the {{Item documentation}} templates to generate queries for problems with the classes but to my knowledge that’s all. author TomT0m / talk page 16:33, 7 December 2022 (UTC)[reply]

. author TomT0m / talk page 11:53, 7 December 2022 (UTC)[reply]

Strong support "value hierarchy property" is a horrible label ... "transitive over" is much better. --Push-f (talk) 07:45, 15 December 2022 (UTC)[reply]

Property talk:P6609

Documentation

Use to indicate transitivity?[edit]

Naming ?[edit]

Navigation menu

Search