Wikidata:Property proposal/precedes/follows various things

precedes/follows various things

appears before phonological feature

Originally proposed at Wikidata:Property proposal/Lexemes

   Done

Description	this lexeme form appears only when another lexeme form with this phonological feature comes after it
Data type	Item
Domain	anything phonological in nature, really
Example 1	e (L640143-F1) → advanced tongue root (Q110629567)
Example 2	a (L640143-F2) → retracted tongue root (Q110629568)
Example 3	অ (L710941-F1) → initial consonant (Q121301466)
Example 4	অন্‌ (L710941-F2) → initial vowel (Q121301470)
Planned use	move phonological feature items in strange places (as grammatical features on forms, as requires grammatical feature (P5713) values, etc.) to this property instead

appears after phonological feature

Originally proposed at Wikidata:Property proposal/Lexemes

   Done

Description	this lexeme form appears only when it comes after another lexeme form with this phonological feature
Data type	Item
Domain	anything phonological in nature, really
Example 1	لر/լար/lar (L941749-F1) → back vowel (Q853589)
Example 2	لر/լէր/ler (L941749-F2) → front vowel (Q5505949)
Example 3	sto (L411808-F1) → Finnic back vowel (Q121301347) (per w:en:Finnish_phonology#Vowel_harmony "In the case of compound words, the choice between back and front suffix alternants is determined by the immediately-preceding element of the compound")
Example 4	stö (L411808-F2) → Finnic front vowel (Q121301348)
Planned use	move phonological feature items in strange places (as grammatical features on forms, as requires grammatical feature (P5713) values, etc.) to this property instead

appears before lexeme form

Originally proposed at Wikidata:Property proposal/Lexemes

   Done

Description	this lexeme form appears only when this other specific lexeme form comes after it
Data type	Form
Domain	any lexeme form
Example 1	내 (L246-F2) → 가 (L740723-F1)
Example 2	제 (L745330-F2) → 가 (L740723-F1)
Example 3	"-co-" (pharmacological affix) → "-xi-", "-zu-"
Example 4	"-me-" (pharmacological affix) → "-xi-", "-zu-"

appears after lexeme form

Originally proposed at Wikidata:Property proposal/Lexemes

   Done

Description	this lexeme form appears only when it comes after this other specific lexeme form
Data type	Form
Domain	any lexeme form
Example 1	라도 (L747165-F4) → 더 (L747172-F1) plus about six other forms
Example 2	"-nü" (relativizing clitic in Ye'kuana (Q3082027)) → "-anö" (past imperfective suffix)
Example 3	"-ggy" (subjunctive/imperative suffix in Hungarian) → "hi" (root of the verb "hisz")

Motivation

These properties are intended to enhance our ability to indicate the conditions under which certain forms may appear—beyond what is currently only possible with the rather scope- and value-restricted P6712 (P6712). (If these proposed properties get created, I intend to promptly nominate P6712 for deletion.) Mahir256 (talk) 01:26, 30 January 2023 (UTC)

Discussion

•  Question This brings to mind an earlier discussion of mutation. @Mahir256: Can you say how your proposals might be related or analogous to that, if at all? ―BlaueBlüte (talk) 05:20, 1 February 2023 (UTC)
  • @BlaueBlüte: If I understand Celtic mutations correctly, I could see the two phonological feature properties proposed applying to a variety of function word forms across languages that trigger certain mutations: "daou" "precedes phonological feature" "soft mutation". (In this circumstance, the existing soft mutation forms on nouns/adjectives/etc. would still need to indicate that they express this phonological change, but that indication would not use any of these proposed properties.) Mahir256 (talk) 05:52, 1 February 2023 (UTC)

    @Mahir256: Okay, thanks, this sounds like it would be helpful then, but see further question/comment below. ―BlaueBlüte (talk) 17:40, 14 February 2023 (UTC)

•  Support for all - thanks for a clear exposition on the need for this ArthurPSmith (talk) 21:05, 3 February 2023 (UTC)
•  Support all propositions ―Eihel (talk) 04:31, 7 February 2023 (UTC)
•  Question Might there be cases in which, when using the proposed phonological-feature properties, there would be a need to distinguish where within the preceding/following utterance the referenced feature appears? For example, suppose for some language a grammar or phonology exists that uses the same category to describe certain qualities of the beginnings and ends of syllables or forms. One could easily see how those qualities might be described using ‘has grammatical feature’ (at least in the current state of affairs of lexicographical data), regardless of where in the form they actually appear. But for the properties proposed here it would seem to be necessary at least in some cases to specifically match the occurrence of features in, say, the anlaut.
  Before deciding on this property proposal I think it could be good to get a sense of how it would fit into these wider considerations. A couple of options readily seem to suggest themselves:
  • The phonological-feature-matching properties proposed could be complemented by feature-declaring properties, possibly such that those declaring properties specify the location of the declared feature within a form, e.g., ‘has phonological feature in anlaut’ (= ‘has initial phonological feature’). A text-generation algorithm could then use a heuristic of matching the value of a ‘matching property’ with the value of the ‘closest’ ‘declaring property’ on the preceding/following form. (Though I’m not sure this would be linguistically sound—does something like ‘alliterative prosody’ exist, just to name a conceivable counterexample?) Or we might think about introducing qualifiers to match not only the feature but also the ‘declaring’ property with which it is stated on forms that might act as preceding/following form.
    • Or, same as above, but the proposals under consideration here would be split into, say, ‘follows phonological feature in preceding auslaut’ (Pxxx), follows phonological feature in preceding root (Pyyy), …; and similarly with the ‘precedes’ property (one result of such a split would then presumably be (close to) the phonological equivalent of P6712 (P6712)).
  • The items representing phonological features could be designed such as to not only represent the feature but also the location of its occurrence, say ‘fricative anlaut’ or ‘back vowel in root’. (This, of course, might considerably bloat the number of items to be maintained.)
  ―BlaueBlüte (talk) 17:40, 14 February 2023 (UTC)
  • @BlaueBlüte: It was my view when writing this proposal that at this time it is probably easier to have dedicated items for different phonological environments ("consonantal anlaut", "high vowel nucleus", etc.; the third of your bullet points above), and that if in fact some pattern begins to emerge among the items that are created, more specific properties as you propose in the first/second bullet points above can be revisited later. Mahir256 (talk) 15:11, 15 February 2023 (UTC)

    @Mahir256: I don’t have any prior knowledge about Finnish phonology, but my guess is that your examples #3 and #4 from “follows phonological feature in speech” wouldn’t work the way you just described. Looking at kirja (L6874) as candidate to which to append one of the suffixes forms from examples #3 and #4, kirja (L6874) would seem to have both a back vowel (Q853589) and front vowel (Q5505949). If that’s the case, then with your example statements sto (L411808-F1)‘follows …’ (Pxxx)back vowel (Q853589) and stö (L411808-F2)‘follows …’ (Pxxx)front vowel (Q5505949) we wouldn’t be able to infer that the correct suffix form is sto (L411808-F1) to give kirjasto (L6795). But I’m guessing the example could still work if the phonological context of the back vowel (Q853589) in kirja (L6874) were taken into account using something like ‘back vowel in final syllable’ (Qxxx) (which might be a new item). Does that make any sense? Could you amend the example as needed? Thanks, BlaueBlüte (talk) 22:56, 22 February 2023 (UTC)

    "i" is neutral in Finnish vowel harmony. - Nikki (talk) 13:14, 29 April 2023 (UTC)

  If there are cases where you need to distinguish whether it applies to the anlaut or the root, you could use a qualifier. - Nikki (talk) 13:07, 29 April 2023 (UTC)

  @Nikki: You're absolutely right about the use of qualifiers, and in principle I would expect that they would be much better in many circumstances than dedicated items, but to reduce (increase?) @BlaueBlüte:'s confusion I've struck the words 'in speech' and 'precedes/follows' from the content of the proposal boxes above and modified the values to use more specific dedicated items in any case. Mahir256 (talk) 17:57, 4 July 2023 (UTC)

•  Comment I think it might help avoid confusion down the road if the properties were labeled more clearly, especially what is now “precedes lexeme form” and “follows lexeme form”, for the sense of direction implied in ‘precede’ and particularly ‘follow’ might intuitively be just as plausibly interpreted etymologically, in terms of agreement, etc. I’m not sure how exactly it would be done (“triggered”? “in speech”?), but clarifying that this pertains to the course of speech would seem to be in order. ―BlaueBlüte (talk) 17:40, 14 February 2023 (UTC)
  • @BlaueBlüte: Noted, and added 'in speech' to the proposed property labels. Mahir256 (talk) 15:11, 15 February 2023 (UTC)

    I find the labels confusing now that they have "in speech" in them. Follows/precedes form in speech... as opposed to in writing? If you really think they're likely to be misused despite the descriptions, I would suggest something more like "used before/after". - Nikki (talk) 12:53, 29 April 2023 (UTC)

    @Nikki: Done; @BlaueBlüte:, now that you're back, can you respond to Nikki's comments from 29 April? Mahir256 (talk) 17:54, 4 July 2023 (UTC)

    @Nikki, @Mahir256: Yes, it seems one would have to come at this with a rather specific linguistic understanding of “speech” for that to work well as labels. The “before/after” solution seems fine, but I would suggest the more descriptive “appears before/after”. ―BlaueBlüte (talk) 18:19, 8 July 2023 (UTC)

    @BlaueBlüte: How about now? Mahir256 (talk) 20:25, 8 July 2023 (UTC)

•  Support with the caveat that I think these should be usable as qualifiers on pronunciation related properties such as IPA transcription in addition to as main values. I forsee this being useful for representing information about tone sandhi rules which condition pronunciation changes on a single form rather than distinguishing forms (see Wikidata:Lexicographical_data/Sandhi_rules#Punjabi for specific examples) -عُثمان (talk) 18:09, 7 March 2023 (UTC)
  • @عُثمان: Completely understandable, and I would agree with this use as well. Mahir256 (talk) 17:54, 4 July 2023 (UTC)
• @Mahir256, BlaueBlüte, ArthurPSmith, Eihel, Nikki: Good discussion. It seems there is no opposition and that the proponent has addressed the main concerns, so I am creating the 4 properties as appears before phonological feature (P11950), appears after phonological feature (P11951), appears before lexeme form (P11952) and appears after lexeme form (P11953). As my knowledge about lexemes is limited, I kindly ask others to help fill the details. TiagoLubiana (talk) 18:00, 14 August 2023 (UTC)