Wikidata:Property proposal/nominalised item for this sense

nominalised item for this sense

Originally proposed at Wikidata:Property proposal/Lexemes

   Not done

Description	concept corresponding to the 'nounified' version of this non-noun sense
Represents	nominalisation (Q1500667)
Data type	Item
Domain	Sense
Allowed values	Senses of Lexemes of a non-noun grammatical category (e.g. verbs, adjectives, adverbs, etc.)
Example 1	white (L3272-S1) → white (Q23444)
Example 2	run (L279-S1) → running (Q105674)
Example 3	grateful (L10686-S1) → gratitude (Q2728730)
Example 4	thankful (L45838-S1) → gratitude (Q2728730)
Planned use	To start using this property on non-noun senses as a way of connecting them to items, like has been done for nouns with item for this sense (P5137).
See also	Wikidata:Property_proposal/adjective_of, item for this sense (P5137)

Motivation

We can currently connect noun senses centrally using item for this sense (P5137). This seems like a way to accomplish the same thing for non-noun senses, since the community doesn't like the idea of adding items for non-noun senses. The sense framework is definitely missing something like this. (As a side note, although 'nominalisation' is a morphological transposition, this property refers to it in more of a semantic light, as it is between senses and concepts.) Liamjamesperritt (talk) 01:18, 8 May 2019 (UTC)

Discussion

@ArthurPSmith, Micru, KaMan, Rua, VIGNERON, Jura1: This is a follow-up idea based on Wikidata:Property_proposal/adjective_of, but just for a broader set of use cases. Since you all had opinions about that proposal, what your thoughts on this? Liamjamesperritt (talk) 01:18, 8 May 2019 (UTC)

I think this is a reasonable approach - would you expect to use this to link the main sense of knit (L15638) with knitting (Q6033694) as a verb example? ArthurPSmith (talk) 02:00, 8 May 2019 (UTC)

Oh I just noticed you had "run" and "running" - that works for me! ArthurPSmith (talk) 02:01, 8 May 2019 (UTC)

 Comment @Liamjamesperritt: thanks for this proposal, I like this proposal better than the previous one (my second remark has now been addressed). My first remark from last time still stands: why having two mutually exclusive properties? I would just use item for this sense (P5137) for each and every lexemes, if the lexeme is not a noun them "nominalisation" can easily be inferred, I don't see the need to put it explicitly in the property. Cheers, VIGNERON (talk) 07:54, 8 May 2019 (UTC)

@VIGNERON: I agree that just using item for this sense (P5137), the context should be enough to differentiate whether it is for "equivalence" or "nominalisation". However I feel that the semantics of item for this sense (P5137) are already well defined and utilized (Wikidata:Lexicographical data/Statistics#Synonyms) as an "equivalence" relationship. "Nominalisation" is semantically separate from "equivalence", so either we request that the item for this sense (P5137) property definition be updated, or we create a new property that explicates this semantic disjunction. Either way, I personally don't think that item for this sense (P5137) as it currently stands semantically should be used in this way. Liamjamesperritt (talk) 11:30, 8 May 2019 (UTC)

 Question Why do we need a new property, instead of simply extending the domain of item for this sense (P5137) to all Senses? Whether something is a noun is already obvious from the lexical category of the Lexeme. Deryck Chan (talk) 13:07, 8 May 2019 (UTC)

@VIGNERON, Deryck Chan: I think it's better to have a new property so that item for this sense (P5137) can be used relatively reliably to find translations; if it is extended as proposed then different parts of speech would be linked as potential translations of one another which seems problematic. ArthurPSmith (talk) 20:02, 8 May 2019 (UTC)

Agreed. Also, if we use item for this sense (P5137) for nominalisation now, and then in the future the community decides to start adding items for verbs/adjectives (which I hope we one day will), then all uses of item for this sense (P5137) from verb senses to noun items, for example, may have to be undone and redirected to the correct verb item. Having a separate property reserves item for this sense (P5137) as a pure equivalence relation (for synonymy and translation). Liamjamesperritt (talk) 23:20, 8 May 2019 (UTC)

@ArthurPSmith, Liamjamesperritt, Deryck Chan: I agree with you, the lexical category matter, for many different cases (including but not limited to translations, it could be synonyms/antonyms/hyperonyms/hyponyms/holonymes/meronyms/etc.) we need the lexical category. But the lexical category is already always indicated so a query (or any other tool) can already easily give the results for a same lexical category (and same goes if we need to move/undo/correct linking to items). My point, is that the Lexemes will quickly be very heavy already, I don't want to duplicate data unnecessarily (per Occam's razor (Q131012)). Cheers, VIGNERON (talk) 08:29, 9 May 2019 (UTC)

@ArthurPSmith: I'm not sure a new property is needed to achieve this reliability. To use item for this sense (P5137) to find direct translations, we need to reverse-query the property anyway (L-yyy links to Q-xxx, which is linked from L-zzz), so we need to build a script or a query anyway. The lexical category can be built into the query. The need to make direct translations reliably transcludable is a good argument to keep translation (P5972) though. Deryck Chan (talk) 09:44, 9 May 2019 (UTC)

@VIGNERON, Deryck Chan: Then should we start a new thread on the item for this sense (P5137) discussion page about broadening the domain of the property? Liamjamesperritt (talk) 11:43, 9 May 2019 (UTC)

@Liamjamesperritt: yes whe should, it would be clearer and stronger after a discussion there. Cheers, VIGNERON (talk) 11:47, 9 May 2019 (UTC)