Property talk:P5972
Documentation
sense (meaning) of a lexeme in another language that corresponds exactly to this sense (meaning) of the lexeme
| Represents | translation (Q7553) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Data type | Sense | |||||||||
| Domain | Sense (note: this should be moved to the property statements) | |||||||||
| Example | According to this template:
When possible, data should only be stored as statements
| |||||||||
| See also | item for this sense (P5137) | |||||||||
| Lists |
| |||||||||
| Proposal discussion | Proposal discussion | |||||||||
| Current uses |
| |||||||||
| Search for values |
List of violations of this constraint: Database reports/Constraint violations/P5972#Conflicts with P279, search, SPARQL
List of violations of this constraint: Database reports/Constraint violations/P5972#Symmetric, SPARQL
|
This property is being used by: Please notify projects that use this property before big changes (renaming, deletion, merge with another property, etc.) |
Translation as property and query edit
Hi,
Do we really a property for translation? More exactly what is the difference between :
- LXXX P5972 LYYY
and
- LXXX P5137 QZZZ
- LYYY P5137 QZZZ
Point-blank I don't see a difference, this seems exactly equivalent to me and the second method seems better to me (as we need to store item for this sense (P5137) anyway). At the very least, maybe we can add some constraints based on this equivalence.
@Tubezlob, Mfilot, Infovarius, KaMan, Jura1:
Cheers, VIGNERON (talk) 07:53, 8 November 2018 (UTC)
- @VIGNERON: That's true as long as we agree to create items for all parts of speech. There seems similar discussion at Wikidata_talk:Lexicographical_data/Archive/2018/11#Translations. KaMan (talk) 08:05, 8 November 2018 (UTC)
- Ping @Tubezlob, Denny, Sintakso, Duesentrieb: as users who requested it.
- I think it all depends on how one uses it. For experiments, it can be good to list all translations for "January", but overall 3 variants between the same language pair could be more interested. Personally, I rather see a statement than a translation in the "gloss" field.
- Furthermore P5137 statements make it hard to view the other lexeme. Obviously, the GUI on all these could use some more development. --- Jura 08:15, 8 November 2018 (UTC)
The assumption is that translations are not necessarily transitive, and that they don't always form a mutually translatable perfect group of Lexemes (in which case the star-based method off item for this sense (P5137) would be better). I do agree that using translation (P5972) when we already have item for this sense (P5137) is problematic and leads to all the problems that interwiki maintenance had previously - I would preferably say to use translation (P5972) only if item for this sense (P5137) is problematic. (Just as we can still can have explicit interwiki links in the Wikipedias). Does this make sense? --Denny (talk) 15:28, 8 November 2018 (UTC)
Kept edit
This property was nominated for deletion and kept. Multichill (talk) 20:32, 27 May 2019 (UTC)
Automatically fill language of work or name (P407) edit
All of the things that use this item should have a language. However, it doesn't automatically show the language of the translation, and it needs to be added as a qualifier. Is there any way that the qualifier language of work or name (P407) could automatically be added to this property and display the language of the other item? DemonDays64 | Talk to me 23:55, 9 January 2020 (UTC) (please ping on reply)
Large translation networks and P5137 edit
There are a number of lexeme senses with quite a few translations on them: while this upper bound is around 20 now, at one point it was considerably larger for some senses (most pertaining to the concepts of "liquid water" and "mother"). This often happens alongside one of the following:
- All nodes in the resulting network have the same item for this sense (P5137) value. This is merely redundant, given that the connection to the Wikidata item in question is the same across all senses in the network. (If the connection was different for some of the senses, that might be an argument for maintaining this property on those senses only and either revising or removing the corresponding P5137 links.) In the aforementioned large sense graphs, this situation had occured with the following:
- All nodes in the resulting network are connected to one another (i.e. it is a complete graph (Q45715)). This is simply not scalable, given that the number of edges in the network would be forced to scale quadratically in the number of languages involved rather than linearly. Moreover, since most of the senses in such networks do not have any other special semantic features, the connections between language senses can be preserved with fewer P5972 links so long as it is possible to follow a P5972 path between them to get from one language to another (the
wdt:P5972*predicate in the Wikidata Query Service). This leads then to the following: - There is one node to which all other nodes are connected, and none of these other nodes are otherwise directly connected to each other. This may have been borrowed from the practice on the non-linked-data-based Wiktionary of, for example, placing a list of translations in other languages for a Dutch word on the Dutch Wiktionary page for that word. Not only does this completely neglect the power of linked data to maintain these translation connections without such exhaustive enumeration, it may also end up promoting the idea that there are "lexicographic centers" from which meaning in other languages emanates—something which has already constricted the diversity of the lexemes added for some languages.
To address the latter two points above with respect to the graphs for "mother", "liquid water", and "house", I had reorganized these graphs so that instead of a single focus or a complete graph, the connections using this property had some sort of logic to them; the connections were chosen based first on genealogical relationship (Paiwan ↔ Malay, Latin ↔ Sicilian), then on geographical proximity (German ↔ Polish, Malayalam ↔ Sinhala), and last on other relevant historico-linguistic relationships (English ↔ British Sign Language, Danish ↔ Kalaallisut). During these reorganizations, the connectedness of the resulting graph was maintained; it is still possible to get from the Finnish "talo" to the Occitan "ostal" through following P5972 paths in the network, and yet this is achieved with fewer direct connections in the overall graph. If there was a sense that got lost in these reorganizations, this was no detriment either; for all three of these graphs, all senses within them had the same P5137 value, and this P5137 value was never removed, so in principle there is still a connection between those lost senses and the rest of the P5972 graph. In fact, if someone really wanted to remove all P5972 connections between senses which both 1) had the same P5137 value and 2) lacked any other semantic properties which might otherwise distinguish them, this would not change the fact that the P5137 connection still exists and could be used to identify equivalent senses for them.
Recently part of such a reorganization for the concept of "dog" was reverted by @Andreasmperu:. I would thus like to see what others think of what has been done and whether the principles which have guided it should persist. Mahir256 (talk) 05:20, 30 January 2022 (UTC)
- If this property is going to have a limit on the number of values, we'll need some guidelines. Verbs, adverbs, adjectives, prepositions tend to have no corresponding item, so no item for this sense (P5137) to help with the connections. Andreasm háblame / just talk to me 05:46, 30 January 2022 (UTC)