Wikidata:Property proposal/translation

From Wikidata
Jump to navigation Jump to search

translation[edit]

Originally proposed at Wikidata:Property proposal/Lexemes

   Done: translation (P5972) (Talk and documentation)
Descriptionword in another language that corresponds exactly to the meaning of the lexeme
Representstranslation (Q7553)
Data typeSense
DomainSense
Example
  • dog (en) → chien (fr)
  • chien (fr) → dog (en)

Motivation

I think that it is possible to put a symmetric constraint (Q21510862) and maybe a new type of constraint: different language constraint. Tubezlob (🙋) 06:58, 17 April 2018 (UTC)[reply]

Agree on the different language constraint, not sure about the symmetric constraint though. --Denny (talk) 17:48, 17 April 2018 (UTC)[reply]

Discussion

 Comment This solution is not scalable because you would need 227 links in each of the 227 lexemes that have a translation for the concept "dog", that is at least 50k statements... It is simpler to link the sense to the item with "equivalent concept" and let the software find out all the translations by itself.--Micru (talk) 07:13, 17 April 2018 (UTC)[reply]

@Micru: But in my example, there is a linked Q-item (dog (Q144)) but it is not the case for the majority of lexemes. I agree that it is a lot of duplicated information, but I don't think there will be more than 200 translations for each lexeme (it is the same issue in Wiktionary).
Should we move the datatype from lexeme to sense too? Tubezlob (🙋) 07:30, 17 April 2018 (UTC)[reply]
@Tubezlob: Ok, there might be cases where it is useful. Yes, it should be sense too. Updated.--Micru (talk) 07:35, 17 April 2018 (UTC)[reply]
@Tubezlob: we should be ready for 6000+ links (according to number of languages). --Infovarius (talk) 09:00, 24 September 2018 (UTC)[reply]
@Infovarius: As stated by Denny, the current Wiktionaries do that. And on top of that there is currently an item with 8343 statements. Tubezlob (🙋) 15:02, 24 September 2018 (UTC)[reply]

 Support It is scalable - the current Wiktionaries do that :) But maybe we should refrain from using the translation property when there is an equivalent concept this connects to? This is going to be fun figuring out. --Denny (talk) 17:48, 17 April 2018 (UTC)[reply]

 Comment I strongly prefer linking words to concepts: in essence, replacing or augmenting the English label 'house cat' of house cat (Q146) with some structured relationship to one sense of the lexeme 'house cat'. In other words, if you want the English label for concept Q146, you'd look for the lexeme with sense->Q146 and language->English. This would also cover synonyms because there might be multiple lexemes with sense Q146, and translations because there might be multiple languages with sense Q146. Runner1928 (talk) 21:13, 25 April 2018 (UTC)[reply]
Update: item for this sense (P5137) now exists for this purpose. Runner1928 (talk ) 20:14, 14 May 2018 (UTC)[reply]
 Comment "indirect" translation relationships via concepts may work for nouns, but they don't work well at all for other kinds of lexemes. Forcing all translation relationships to go via concepts (Items) would lead to an explosion in the number of items, if you want to cover things like "cosy" and "quick" (as opposed to "fast" as opposed to "swift"?), or things like to "to warble". But even four nouns it gets tricky... how about words "outrage" or "nuisance"? Trying to connect such words in different language via a neutral concept would lose all nuance. I agree though that indirect translations via a "denotes" relationship is useful; perhaps we can have both kinds of translations shown together in the UI some day. -- Duesentrieb (talk) 18:52, 12 June 2018 (UTC)[reply]

 Support I believe that distinct property is the best way to indicate translations. The concept linking approach might work well for nouns, but I doubt it would be useful for other parts of speech. --Sintakso (talk) 06:18, 24 May 2018 (UTC)[reply]

On hold ready for creation once datatype available.
--- Jura 12:49, 12 June 2018 (UTC)[reply]