Wikidata

Wikidata:Property proposal/item for this sense

< Wikidata:Property proposal

item for this sense edit

Originally proposed at Wikidata:Property proposal/Lexemes

Descriptionconcept corresponding to this sense of a lexeme
Data typeItem
DomainSense
Example
  • dog-sense-1 (domestic animal) (en) → dog (Q144)
  • chien-sense-1 (animal domestique) (fr) → dog (Q144)
See alsotranslation

Motivation

Following User:Micru's suggestion above, this seems essential (in fact I'm a little surprised it's not a central part of the Lexeme data model?) ArthurPSmith (talk) 12:51, 17 April 2018 (UTC)[reply]

Discussion

  • @Denny: good question. I guess I was imagining this as an assertion that this is the closest wikidata item to the meaning. Since puppy (Q39266) exists, en:puppy would point there, not to Q144. Similarly for your other examples which seem to have their own items. However, both en:pig and en:swine would point to pig (Q787) as they are (in English?) interchangeable names for the same animals. ArthurPSmith (talk) 19:08, 17 April 2018 (UTC)[reply]
Just to see if I understand correctly: so if we wouldn't have a topic for puppy, it would be ok to link en:puppy to the dog topic? --Denny (talk) 19:34, 17 April 2018 (UTC)[reply]
Thinking about this more, and I have switched from a support to an oppose. What if we, instead had three properties - based on the SKOS model - for "exact match", "narrower than", "broader than"? I.e. "everything this sense correctly and literally refers to is within this item" for exact, etc.? So we would say "en:puppy" narrower than Q144, "en:dog" exact match Q144, etc.? --Denny (talk) 20:06, 17 April 2018 (UTC)[reply]
@Denny: we have mapping relation type (P4390) as a qualifier if necessary for cases like that. Can you think of an example where it would be really needed though? ArthurPSmith (talk) 12:35, 20 April 2018 (UTC)[reply]
I just gave examples, like en:puppy. By using a qualifier for such a basic difference we are complicating the querying considerably. I think with the three suggested properties we would have a simpler solution(well, there might be also a "somehow overlaps" relation, though). --Denny (talk) 16:05, 20 April 2018 (UTC)[reply]
That's not a legitimate example as puppy (Q39266) exists. ArthurPSmith (talk) 17:06, 20 April 2018 (UTC)[reply]
Ah, understood. The English word "Actress", or basically all professions with a male and female form; English "Elephant bull" or terms like "billy" or "kid" for male goats or child goats respectively; the different pieces of beef cuts, which are all named, but only about half of them have items, like the Portuguese Alcatra; in general, the item space about common nouns is by far not as detailed as the lexeme space in almost any language - never mind that it sometimes is rather complicated. Where would you let the English word "mouse" point to? I don't think any of our items is exactly right. The difference between a carpenter, the field of carpentry, the products of carpentry, etc.
But you wanted one example, let's stay with 'kid' in the sense of little goat. --Denny (talk) 16:31, 21 April 2018 (UTC)[reply]
Which seems to be covered by goat kid (Q13048847) (Spanish "cabrito"). I'm not saying we necessarily need an item that exactly matches every sense of every word in every language, I don't think that's going to be either possible or a good idea. But we have two choices here - either add links that any reasonable person would say are the same thing and not add links to items if there is no such "same" item (without worrying overly about the degree of "exactness" of the relation), or allow links when it is clear they are not the same with either a qualifier or distinct property in such cases. Is it really useful to link 'kid' to goat (Q2934)? Perhaps kitz (Q1743952) is closer? I don't see that these "less than same" links would be particularly useful due to their ambiguity (what is the parameter in which they differ?); this proposal in any case was intended to only do the first piece, to add links when the item really does correspond to the sense by any reasonable understanding. ArthurPSmith (talk) 13:19, 23 April 2018 (UTC)[reply]
If it is meant to point to corresponding items only - not to the closest item - then I will switch immediately to a support again. I think for the other terms we can always use hypo-/hpernym relations to other words and have those then point to the items using this property. So, yes to 'corresponding', no to 'closest'. --Denny (talk) 16:17, 23 April 2018 (UTC)[reply]
Ok, that makes sense to me. I believe I used "corresponding" in the original description, not sure anything needs to change in the wording of the proposal? ArthurPSmith (talk) 19:42, 23 April 2018 (UTC)[reply]
Nope, all good. --Denny (talk) 19:52, 23 April 2018 (UTC)[reply]
Retrieved from "https://www.wikidata.org/w/index.php?title=Wikidata:Property_proposal/item_for_this_sense&oldid=683986005"