Template talk:hy-pron

Western Armenian Pronunciation

When will this template utilize Western Armenian pronunciation? --Lo Ximiendo (talk) 16:11, 14 May 2014 (UTC)

When I have time to learn Western Armenian and add the functionality. Probably in 1 year. --Vahag (talk) 16:15, 14 May 2014 (UTC)

You better hurry... —Aryamanarora ^{(मुझसे बात करो)} 17:35, 7 May 2016 (UTC)

I'm busy. --Vahag (talk) 17:37, 7 May 2016 (UTC)

It took more than one year, but we can now add support for Western Armenian with @Pompyxmori's help. @Erutuon, is it possible to add an alternative phonetic map to Module:hy-pronunciation for Western Armenian and call it here with a {{a|hy-W|standard}} tag? --Vahag (talk) 06:41, 10 October 2020 (UTC)

@Vahagn Petrosyan: Yeah, I've created sandbox modules and templates to avoid messing up existing uses of {{hy-pron}} and {{hy-IPA}}. {{hy-pron/sandbox}} and {{hy-IPA/sandbox}} use Module:hy-pronunciation/sandbox, which has separate east and west tables for Eastern and Western Armenian. (Both of them are currently just copies of the tables in Module:hy-pronunciation, which I guess are Eastern Armenian.) If you add actual Western Armenian pronunciations to the tables, they can be used in the sandbox templates by setting the system in the module invocation with {{#invoke:hy-pronunciation/sandbox|pronunciation|system=west|word}}. Hopefully this setup will let you get started, but let me know if there are any problems. — Eru·tuon 19:18, 11 October 2020 (UTC)

@Erutuon: thank you very much. I will deal with this over the weekend. --Vahag (talk) 16:38, 12 October 2020 (UTC)

@Erutuon: I added the Western Armenian table, but I don't understand how can I call the module with separate east and west tables in {{hy-pron/sandbox}}. We can change the latter to call Module:hy-pronunciation/sandbox directly, if that simplifies the matter.

Preferably, there should be a way to suppress the display of either Western or Eastern in certain cases, e.g. by using east=off and west=off.

I have added a rule for ['յու'] = 'ʏ' in Module:hy-pronunciation/sandbox which I don't know how to implement.

Finally, I have added the testcase at Module:hy-pronunciation/testcases for multiple word phrases. Can you add the ability of generating the secondary stresses for all polysyllabic words preceding the last word? The rules are the same as for primary stress. --Vahag (talk) 09:13, 24 October 2020 (UTC)

I think I sort of answered all your questions with edits, though I forgot the {{IPA}} around the module output. Changed the module to individually add stress to all space-separated words, using the same code that added stress to the whole input. Added a function replacement that checks the character before յու (yu). — Eru·tuon 10:53, 25 October 2020 (UTC)

@Erutuon: excellent. I have imported your changes into the main templates. But the rules for adding stress and ['յու'] = 'ʏ' do not work. See Ամերիկայի Միացյալ Նահանգներ (Amerikayi Miacʻyal Nahangner) and մյուս (myus).

Is there a way to call Template:hy-IPA with a system=west parameter? --Vahag (talk) 11:36, 25 October 2020 (UTC)

I've changed Module:hy-pronunciation so that it takes |system= either in the module invocation or in {{hy-IPA}}. — Eru·tuon 22:19, 25 October 2020 (UTC)

@Erutuon: great, it works! But ['յու'] = 'ʏ' in for Western Armenian still doesn't work. E.g. մյուս (myus) should be [mʏs] for Western Armenian. --Vahag (talk) 05:47, 26 October 2020 (UTC)

@Vahagn Petrosyan: Aha, figured it out. It was an ordering issue: ու (u) was being processed before յու (yu) because the order of iteration for string keys in a Lua table is unpredictable. Changed the structure of the table. — Eru·tuon 21:32, 26 October 2020 (UTC)

@Pompyxmori, @Hovsepig, is the initial [ə-] in words like ստեղծել (steġcel), սկիզբ (skizb), սպանդ (spand) obligatory or can it be dropped? If it can, are the ə-less forms considered informal? In Eastern Armenian, ə- is nowadays almost always dropped, both in formal and informal speech. --Vahag (talk) 14:18, 26 October 2020 (UTC)

@Vahagn Petrosyan: I believe they are obligatory, both formally and informally. In fact, it's difficult for me to pronounce them without the schwa. (The pronunciation reminds me of Russian consonant pairs). User:Pompyxmori (talk) 16:13, 26 October 2020 (UTC)

@Pompyxmori: I see. What do you mean the pronunciation reminds you of Russian consonant pairs? You mean ə-less pronunciation? --Vahag (talk) 17:25, 26 October 2020 (UTC)

@Vahagn Petrosyan:. When I said consonant pair, I should have used the term "consonant cluster". For example, the word սցենարիստ. This consonant cluster is hard to pronounce for Western Armenian speakers. Just like the monophthongs «իւ» or «էօ» for Eastern Armenian speakers.

@Vahagn Petrosyan:. Yes, the schwa is obligatory before a {s/z}{p/t/k/b/d/g} cluster in Western Armenian. Sometimes it is optionally dropped in connected speech [gadu (ə)ste...] կատու ստեղծեցի. But in citation forms, the schwa is obligatory. Inside a compound, the schwa is inserted if the second-member is the meaningful head: բանաստեղծ but տարեəսկիզպ.Hovsepig (talk) 20:22, 26 October 2020 (UTC)

I see. So it is obligatory in Western Armenian but prescriptive in Eastern Armenian. @Erutuon How can I give different rules for east and west at the stage where the letters are IPA symbols? --Vahag (talk) 06:23, 1 November 2020 (UTC)

@Hovsepig, User:Fenakhay has kindly added a parameter for suppressing Eastern or Western Armenian like this and this. Please use the feature. --Vahag (talk) 09:57, 13 January 2021 (UTC)

@Vahagn Petrosyan: nice feature. It shall be used. I managed to update around 1k words to hy-pron last weekend, and I’ll try to do a 1k more this weekend. Thankfully I’ve had few occasions of needing to suppress the WA form Hovsepig (talk) 10:10, 13 January 2021 (UTC)

Stress marker inside optional glide

Moved from Module talk:hy-pronunciation.

The algorithm incorrectly adds the stress marker inside an optional glide: թի(յ)ակ --> ti(j')ak https://en.wiktionary.org/w/index.php?title=%D5%A9%D5%AB%D5%A1%D5%AF&action=edit Hovsepig (talk) 07:01, 25 October 2020 (UTC)

Erutuon fixed it. --Vahag (talk) 05:47, 26 October 2020 (UTC)

Thank you kindly! Hovsepig (talk) 20:22, 26 October 2020 (UTC)

round vowels in diphthongs

@Vahagn Petrosyan @Erutuon I see that the module pronounces <յու> as [Y] for Western. But I don't think this is fully accurate. Most words with <յու> are variably pronounced as [ju], [ju], [Y]. Some tendencies 1) The suffix -ություն is generally pronounced with [Y], and so are a few words like գյուղ (gyuġ) and հյուսիս (hyusis); but they can optionally take [ju] 2) Most words are just [ju], like առյուծ (aṙyuc), մյուս (myus), ձյուն (jyun), հյուս (hyus), հարյուր (haryur), սյունակ (syunak), նյութ (nyutʻ) etc. 3) Many words in (2) allow an optional [Y] but it varies by speaker and location. I personally never do [Y] for most words in 2; the vowel [Y] is a marginal phoneme for me because I don't speak French)

Proper nouns: Proper nouns like Գժ Error: a bunch of errors are created because of the rule like հայուհի (hayuhi) is [haYhi] where <յու> follows a vowel; same for other words like պայուսակ (payusak)

I see in the first section of this Talk that there was effort in adding the [Y] rule, but I think it should be deactivated. You can add a rule that specifically the suffix -ություն takes [ju]. Perhaps also use a small list for the forced-Y takers like գյուղ; I can go through my Wiktionary scrape and tell you all the members of this list. Except after a vowel or word-initially, all other instances of <յու> is pronounced as [uj]. Hovsepig (talk)

I've made the ⟨յու⟩ handler behave as advertised at least; it was only working word-initially. — Eru·tuon 18:48, 28 October 2020 (UTC)

@Erutuon: thanks! Hovsepig (talk)

@Hovsepig: Erutuon already fixed the obvious error you mentioned. For the other cases, what you say is not necessarily true for all Western Armenians. I have personally heard repatriants from Syria pronouncing հարյուր (haryur) with an ü. Sakapetoyan transcribes all իւ's after a consonant as ü. E.g. from your examples, ձիւն-ակալ, հիւս-կէն, հարիւր-ալիդր, սիւնակ, նիւթ. So, ü is the standard, and {{hy-pron}} generates the standard pronunciation marking it as "Eastern Armenian, standard" and "Western Armenian, standard". I do realize that almost no one follows the prescriptive pronunciation (true also for Eastern Armenian), but handling non-standard pronunciations would complicate matters a lot. I don't want to deal with them yet. --Vahag (talk) 09:36, 29 October 2020 (UTC)

It is true that [Y] a pretty common pronunciation for <յու>, but I would personally treat [ju] as prescriptive and [Y] as colloquial. I didn't know about that Sakapetoyan book, and that threw me into a bit of a hunt in my personal library for Western phonology. Sadly, all the grammars diverge on this point -- Fairbanks (1948) has [Y], Sakayan 2000 has [ju], Hagopian 2007 has [ju], Avetisyan 2011 has [Y], Khanjian 2009/2011 documents his own personal variation which I also don't match 100%. I guest without a definitive grammar of Western, I guess the best approach is to just stick with either [Y] or [ju]. Hovsepig (talk) 15:42, 29 October 2020 (UTC)

necro, but what about իւր? i can agree that other իւԲ words have [ʏ] based on my experiences and what i was taught, but for this word i am pretty sure the default pronunciation has [jʏ] or [ju]: so իւրաքանչիւր should be [j(ʏ/u)ɹakʰanˈʧʏɹ] according to the most cultivated SWA pronunciation. what do @Hovsepig and @Vahagn Petrosyan think, though? which is better? or do you usually hear [#ʏ] here? RagingPichu (talk) 20:22, 1 May 2024 (UTC)

I’ve heard #yr… and #jyr… it varies Hovsepig (talk) 20:42, 1 May 2024 (UTC)

hm i see... seems the dictionary vahagn sent prescribes [ju] at the beginning of a word, but it also says that ր and ռ have merged, which isn't true in all communities, so i'm not sure how reliable that is. i will ask some friends for their thoughts as well RagingPichu (talk) 21:59, 1 May 2024 (UTC)

I don't know. The only Western Armenian I know is my Syrian Armenian barber Avo Shadoyan, who says [hæ'ɾʏɾ]. I haven't paid attention to other words containing իւ. Vahag (talk) 11:26, 2 May 2024 (UTC)

hm... by any chance, is he from northwestern syria? or is his family from urhay or thereabouts? this reminds me of the extreme south and southwestern dialects of armenian, who pronounce ա as [æ] by default—it is not SWA. from what i've heard, many many syrian armenians hail from cilicia. anyway, speaking with other people has brought up more issues, since a lot of people just don't use [ʏ] at all and say [ju] instead... but i can say that no one pronounced #իւ without a [j], so i will amend the pronunciation to have [jʏ] for now and keep my ears peeled for իւղ, իւր(աքանչիւր), եւ այլն։ and i am (barely) acquainted with someone who speaks extremely good, fluent, literate western armenian, so i will ask him and see whether he responds. edit (13:47): it seems that յուր (yur) has a ZWSP there, and is the source of the odd pronunciation: the others have [ju]. looks like i kicked up a fuss over nothing, my apologoes RagingPichu (talk) 13:39, 2 May 2024 (UTC)

nasal assimilation

@Erutuon @Vahagn Petrosyan The code will transcribe a cluster <նխ> into [ŋχ]. However, I can't find documentation in the phonological literature for this assimilation rule. Vaux 1998:16 and Khachatrian 1988:180 document nasal assmilation of ն to ŋ before <կ,ք,գ>, but not before խ. Do you agree with Khachatrian's reports that <նխ> is transcribed as [nχ], or does your EA speech do [ŋχ]?

Also, both sources document that ն becomes [m] before bilabials: անբայման [ampajman]. Your recording for անպայման applies assimilation while անբիծ does not. Though անբիծ sounds like the speaker is actively placing places between every syllable instead of speaking naturally.

Hovsepig (talk) 01:47, 3 January 2021 (UTC)

@Hovsepig: I agree that [ŋχ] is wrong. I removed it.

I am not sure if [n] > [m] before bilabials is standard or colloquial, but I indeed found the rule in Khachatryan 1988: 106 and encoded it.

You do not have to ping Erutuon. He is a heavy artillery to be brought out when I cannot code. --Vahag (talk) 07:49, 3 January 2021 (UTC)

@Vahagn Petrosyan how come there's now a rule that introduces ɲ? i don't think i've seen this sound discussed in previous work on armenian Hovsepig (talk) 06:56, 28 November 2023 (UTC)

@Hovsepig The rule was introduced by @Snowman304 without discussion. Snowman, what is this devilry with [ɲ]? Vahag (talk) 18:08, 28 November 2023 (UTC)

@Hovsepig and @Vahagn Petrosyan, that was based on this book: https://web.archive.org/web/20230130032555/https://vahagnakanch.files.wordpress.com/2011/04/modern-eastern-armenian.pdf, specifically pages 19 and 28. If that's a wild interpretation on my or the author's part, I apologize. Snowman304 (talk) 05:15, 29 November 2023 (UTC)

@Snowman304@Vahagn Petrosyan Ah I see. Dum-Tragut refers to Vaux who then refers to Khachaturian 106. But the 106 doesn't talk about palatal assimilation. Page 174 from Khachaturian does and it's translates as:

• Before a palatal sonorant, the lowering of the tongue-middle and then raising (for pronouncing /j/) is done subsequently and so fast that sometimes those two periods approaching creates a palatal n, which is heard like a soft n (like rusisan), like in njard, njut’. While mjus mjasnikjan*

It seems she's talking about how in fast speech a person could pronunce /nd͡ʒ/ as [ɲd͡ʒ]. But this type of phonetic change is so subtle/atypical that it's not the kind that you would see in a transcription dictionary. You could say the same for English (or any language) in that fast speech can make nasals assimilate on a very fine-grained phonetic level (that's hard to hear). The same goes for labial assimilation. It would be best to remove the nasal assimilation rules -- only the velar assimilation is categorical enough (applies 100% of the time) to be a phonological rule Hovsepig (talk) 05:45, 29 November 2023 (UTC)

Colloquial entries

@Vahagn Petrosyan It would be nice if the text for `coll=` used the phrase "(Eastern Armenian, colloquial)" instead of just "(colloquial)"; same for `collw`. Otherwise, it's somewhat odd that both coll and collw return just the text "(colloquial)". People are smart enough to deduce that each "colloquial" phrase is for the separate dialects, but automatic scrapers aren't that smart, and it can cause issues down the line :/Hovsepig (talk) 00:34, 24 January 2021 (UTC)

trilling

the current code includes a rule that changes ɾn to rn (րն to ռն). The rule is applying to ազորներ ալտերնատիվ @Vahagn Petrosyan do you think this rule is needed? Hovsepig (talk) 16:26, 29 November 2023 (UTC)

@Hovsepig, that's a bad rule. It is colloquial and even colloquially does not apply in every case: for example in արնախում (arnaxum) or կտորներ (ktorner) it is always [ɾn]. @Snowman304, please undo your edit. Vahag (talk) 18:31, 29 November 2023 (UTC)

midvowels

Wiktionary uses ɛ,ɔ for the է օ vowels. That's the IPA values used in Vaux. But more recent phonetic work uses /e, o/. It may be wise to change the Wiktionary code too. Hovsepig (talk) 21:01, 1 December 2023 (UTC)

@Hovsepig: Wiktionary's module generates the narrow values [ ], not the broad values / /. So it should be [e̞, o̞], not /e, o/. Vahag (talk) 22:36, 1 December 2023 (UTC)

@Vahagn Petrosyan Ah I see your point. Though even in that case, /ɛ/ is just as broad as /e/. But the difference is that /e/ is relatively more common as a phoneme cross-linguistically than /ɛ/. So if a person sees /e/, they know it’s a broad symbol, while /ɛ/ is more likely to be incorrectly seen as a narrow symbol. Hovsepig (talk) 22:41, 1 December 2023 (UTC)

I made the change. By the way, you may want to add the Armenian to the list of languages in Mid front unrounded vowel and Mid back rounded vowel, referencing the new article. Vahag (talk) 22:57, 1 December 2023 (UTC)

pronunciation of ր

i know that in the literature they write /ɾ/, but considering that the pronunciation of ր according to the prescriptions of standard eastern armenian demands no contact between the tongue tip and the roof of the mouth, and that words like տարր and անդորր exist with a geminate ր, the transcription [ɾ] is not tenable: taps/flaps cannot be geminate by their very nature. i would say that a more appropriate transcription would be [ɹ] (at least for these kinds of words and their derivatives). i will look for spectrographic evidence one way or another for [ɾ] vs [ɹ]. i suspect the transcription [ɾ] arose due to english's customary transcription system, which was established before ɻ as a symbol existed in the IPA, and which eastern armenian ր is qualitatively very different from, plus a few other reasons that would belong in a published article and not on a wiktionary talk page. RagingPichu (talk) 19:26, 26 December 2023 (UTC)

@RagingPichu you can check this thing that I was a part of https://scottseyfarth.com/docs/SeyfarthDolatianGuekguezianKellyToparlak.pdf Hovsepig (talk) 20:27, 26 December 2023 (UTC)

based on this, i would call the spirantized pronunciation primary and the tap secondary, evidenced by the fact that only the former is found in words with a geminate, which warrants a change in the module in my opinion. additionally, the word րոպէ almost always is pronounced with [ɹ] (as far as i could read, it's not mentioned at all in the paper(!), so you'll have to trust me...), which points to me that [ɹ] is the primary realization of /ր/ in RoA eastern armenian. RagingPichu (talk) 20:44, 26 December 2023 (UTC)

@RagingPichu I agree that spirantization is the primary phonetic manifestation of the rhotic. However, one should distinguish phonological representations from phonetic representations. A dictionary cannot encode the entirety of a word’s phonetic manifestation because it can vary by speaker, region, and random chance. A phonological representation is more appropriate as a more abstract representation — and to also maintain consistency with preexisting work on the laminate. So the rhotic should stay as a flap in that regards. And I’d disagree with treating the approximant as default simply because that’s one of the largest difffences between Eastern Armenian vs Iranian Armenian (where ր is an approximant, as cited in https://langsci-press.org/catalog/book/354) Hovsepig (talk) 21:26, 26 December 2023 (UTC)

the primary difference between iranian ր and RoA'an is the place of articulation, not the manner of articulation, as is mentioned in that very book. i wouldn't bring up this topic if armenian wiktionary didn't present everything in phonetic brackets rather than phonemic slashes: indeed, when keeping previous research (despite its incorrectness) in mind and for the sake of backwards compatibility, /tɑɾː/ is the most correct transcription; however, [tɑɾː] is blatantly incorrect and must be [tɑɹː]. RagingPichu (talk) 21:33, 26 December 2023 (UTC)

tangentially, i will say that, even though it is nowadays' practice, holding onto previous research in cases like this perpetuates errors: for example, the claim that eastern armenian generally has ejectives circulated for a few decades after Allen's work in the '50s and has been amended in the field only embarrassingly recently. the same holds for english conventional transcription, where there's a sort of covert conflict between preserving the difference between /ʌ/ and /ə/ in transcription and collapsing these as /ə/ in research focusing on 'standard' american, where this difference is largely not maintained; other similar cases exist as well. these come from a combination of the field not catching up with change of language and of old errors not getting corrected purely from mere convention. RagingPichu (talk) 21:40, 26 December 2023 (UTC)

@RagingPichu regarding Iranian, Vaux would transcribe the Iranian rhotic as ɹ (alveolar approximant) and perceive it as the American R. In my book, I also perceived the Iranian rhotic as an American R. I provided some examples to Seyfarth who perceived it as ɻ (retroflex approximant). Acoustically, the two approximants are quite close and one needs articulatory data to be 100% sure what we hear. But, whether Iranian is ɹ or ɻ (a variant of American R), it's still markedly sounds different from the EA/WA rhotic which sounds like ɾ (an alveolar tap/flap).

Yes, phonetically the flap is spirantized (even so in early work by Allen 1950). And you're right that we use phonetic brackets here. So we have to decide on a level of phonetic detail that's acceptable. Right now, the phonetic detail accepted is schwa insertion and voicing assimilation and velar place assimilation. These processes are relatively more perceivable to the untrained ear than spirantiziation, and such processes are more systematic/categorical than the occassional approximant form in EA (which I have not perceived in WA). Hovsepig (talk) 13:22, 27 December 2023 (UTC)

in that case vaux doesn't know that american /r/ is essentially never [ɹ], or else is ignoring it for the sake of typographic convenience and convention: from what articulatory studies i have read, this consonant is not alveolar but is retroflex or maybe post-alveolar (for those speakers who do not produce the so-called "bunched r"), ignoring the labialization of course. the transcription with ⟨ɹ⟩ is purely customary, a holdover from the early 20th century. i have seen plenty of newer transcriptions using ⟨ɻ⟩. RoA armenian ր is clearly alveolar, while iranian ր is post-alveolar (or retroflex, they sound too similar to me). the transcription /ɻ/ is the correct choice in this year's book released. if you do spectrography comparing "spirantized" RoA ր and iranian, you should see a lower F3 in the latter as the primary difference: this i will say with confidence, and without a source. RagingPichu (talk) 17:45, 27 December 2023 (UTC)

We can show both the phonemic transcription between / / and the phonetic transcription between [ ], like in Finnish ametistihapero. Both can be generated automatically. Our current transcription is a mix of phonemic and phonetic. Vahag (talk) 18:03, 27 December 2023 (UTC)

a very reasonable compromise, and it would be good to separate out phonemic and phonetic transcription more cleanly RagingPichu (talk) 18:08, 27 December 2023 (UTC)

details of the [] pronunciation

so i finally decided to get around separating the phonemic and phonetic represenation, but now remains to decide the degree of precision of pronunciation. the primary decisions to make answer

1. will [ə] be tolerated?,
2. which diacritics to put on which vowels, if any?,
3. how specific to make the consonant transcriptions?,
4. and should the stress mark be used?.

to wit on each question:

• regarding 1., i reference the fact that armenian /ə/ is mid-central in neither dialect, according to that one paper toparlak is involved in, and also protest that it is too vague of a symbol anyway and that better options (viz. [ɜ ɘ]) are available and should be used in a phonetic context, especially in consideration of its exceptional use under stress in turkish loans and onomatopoeiae, requiring more precision, and in consideration of a potential expansion of the wiktionary armenian project to further integrate dialectal data, some of which make richer central vowel distinctions (ee.g. kessab, maragha);
• regarding 2., i would say that [ɑ̈] for SEA is probably reasonable, and then SWA could use that or [ɐ] or [a]; that [e̞] is probably the better choice for both; and that [i u o ʏ] should be left as are;
• regarding 3., i speak specifically on whether to notate the coronal stops as dental [t̪ d̪ t̪ʰ] etc., pronunciations which i have no outside source for except for the confidence of speech pathologist i personally know; on whether SEA ""voiced"" stops and affricates should be notated as [b̤ d̤ ɡ̈] etc., following that one paper seyfarth was involved in (i think so); on whether the mentioned tense voice should be notated, too, since most SEA speakers have some kind of glottalization there (i think so), and if yes, how to notate it, since, as is extremely typical for phonation types, the IPA is sorely lacking there; and on the non-neutralizing devoicing that especially /ɹ r/ but also like /v/ (word-finally, but retaining a lax pronunciation) are subject in certain environments, and how to notate that almost [ʃ]-like quality the first possesses (maybe with [ɹ̝̊] or something...) in words like մարդ and համար;
• and regarding 4., i consider the fact that, strictly speaking, stress is a purely phonological phenomenon, therefore belonging not in the phonetic transcription, and that this is often ignored, so we can choose to keep it or replace it with something more precise like an acute and/or a halflong mark (ambivalent here).

p.s. all foreign languages lacking armenian-style punctuation is their most annoying deficiency. RagingPichu (talk) 16:46, 1 May 2024 (UTC)

I am not good at phonetics, so I can't answer any of your questions (@Hovsepig the Phonetician can), but since you have now separated phonemic from phonetic, it makes sense to give very precise data in [], otherwise it adds little value over //. No general resource gives precise phonetic values for Armenian. If we can do it, Armenian section Wiktionary will become even more exclusive and sexy. Vahag (talk) 15:19, 2 May 2024 (UTC)

I vote against a lot of these changes. First off, I don't exact the same template, but I can gather some of the changes from looking at some word entries

1) the ա is now phonemic /a/ and phonetic [ɑ]

2) the ր is now phonemic /ɹ/ and phonetic [ɹ]

I'm against those changes because they're not accurate from a linguistic angle. The vowel ա is pronounced as a general phonemic back /ɑ/, not as a front [a]. Given that it's the only low vowel in Armenian, then it would have a wide array of possible pronunciations in real life in connected speech, but not the kind of variation that is systematic enough to make it have /a/ has its phonemic form. Same for /ə/. It's the reduced generic vowel; so people will vary in their exact tongue positions depending on the word/sentence/minute, but again, the default transcription is /ə/.

For example, the Toparlak paper that you mention (that I'm in :P) provides a vowel chart where you can see that ա is a back vowel /ɑ/ and not front /a/ because it's back, and also the schwa is mid-central (page 18 https://www.cambridge.org/core/services/aop-cambridge-core/content/view/69DD24FFE4FF3ABE79A5703FCBE345E7/S0025100323000130a.pdf/div-class-title-armenian-yerevan-eastern-armenian-and-beirut-western-armenian-div.pdf)

As for the schwa, the paper even explicitly mentions that it has a mid-central pronunciation in EA (page 20), while seems a bit more wider pronunciation and raised [ə̝]

For the ր, it's pronounced phonetically as a fricative, and rhotics tend to have wide-ranging acoustic manifestations. The Armenian one is typically a tap /ɾ/, and not an approximant /ɹ/ because you can feel your tongue tapping the roof of your mouth. In pronunciation, the tap /ɾ/ will be accompanied by frication (which you can see on a sound wave or spectogram for random ր-final words).

For the voiced stops, yes Yerevan voiced stops tend to be breathy. But that's a regional tendency. One cannot expect all EA speakers to phonetically show breathiness. Same for glottalization; it is often heard for Yerevan speakers, but it is by no means a systematic process. For the most recent study on EA ejectives, Toparlak+me found her specific sample of people glottalized, but we still have a lit review of other studies who didn't, showing the overall free variation of it all https://guarant.cz/icphs2023/15.pdf

For coronal stops, yes, typically they're dental by default. But again, there's free variation in the sense that you can find some speaker saying some word with a coronal pronunciation at least sometimes. Impressionistically, when I was doing my grammar of Iranian Armenian, most of the tokens for coronal stops were felt as dental by my main consultant, but she said some words feeling like it's alveolar. But I just followed the general tendency and literature to treat them as dental.

Now given all this background info, I vote the following:

1) the ա ր should be reverted to what they were before

2) the dental + breathy + schwa changes shouldn't be done

The reasons for my position are thus

1) a narrow phonetic transcription is highly variably by speaker and utterance. So I can pronounce some word դար as dental [t̪ɑɾ] in one context with a final more sonorant tap, maybe more coronal [tɑɾ] in another, maybe more fricationie in another context. So when such super narrow transcriptions are more common to see for individual audio files, not for a generic resource like a dictionary. For example, the schwa is often deleted in connected speech for certain words. Like the word քշող could be pronounced as [kʰəʃoʁ] in one sentence but [kʰʃoʁ] in another.

2) from a utility angle, why would I want my dictionary to include extra diacritics for a phonetic articulation that's variable? By putting both phonetic and phonemic representations, I expect to find a major distinction between them that merits having two levels. For Armenian, the more obvious reasons for a distinct phonemic level is because the Armenian schwa is oftentimes epenthetic, meaning մկրտիչ would have an underlying /mɡɾdit͡ʃ/ while surface [məɡəɾdit͡ʃ]. But that's a pretty abstract phonemic representation (https://phondata.org/index.php/pda/article/view/68) , so it makes sense why Wiktionary shows its epenthetic schwas as /məɡəɾdit͡ʃ/, It's more useful this way too.

3) there should be a precedent for any phonetic 'decisions' on a transcription. That is, a published source (or systematic fieldwork with audio evidence and corpora) should come out and say something like "we found that ր is [ɹ] and not [ɾ]". Otherwise, it's too drastic of a change/argument to be believable.

If you want such phonetic information be known about Armenian, I think it makes more sense to put them in the Wikipedia page on Armenian phonology, but explicitly provide references and citations for them. Otherwise, adding all this variable phonetic info on Wiktionary would decrease Wiktionary's utility.

Side note: For reference, check the Wiktionary entries for English apple and latter. For apple, we see phonemic level /æpəl/ and phonetic [[æpɫ̩] with a syllabic /l/, and latter has phonemic /ˈlæt.əɹ/ with flapping [ˈlæɾ.ɚ] (and the rhoticized schwa). The Appendix:English_pronunciation page explains the logic behind the two levels, and it is well-known in linguistics (with citations) that American English has systematic rules for creating syllabic /l/ and flapping. The phonemic version could be seen as careful speech here, while phonetic is more natural speech. But the point is that the difference between the two English levels is about how there's systematical phonological rules to go from one to the other (and these are well-established rules in the academic literature on English phonology). Hovsepig (talk) 16:51, 2 May 2024 (UTC)

that extra information you mention exists in other languages, like in finnish, or in spanish, or english, or turkish. the extra information in the dictionary is useful for learners trying to refine their pronunciation, for linguistic casuals who spend their free time hunting down "interesting allophony" on wiktionary (this is a real thing, i know people in my undergrad who do this), and just for documentation purposes. in fact, the finnish, which is the most narrow of the examples i gave, is the example vahagn brought up. this is why i bothered to separate the phonemic from the phonetic transcription in the first place.

the reason i did /a/ is because i have in fact seen it before. it might've been in the older literature to be fair. anyway, no one actually uses that symbol to mean the front vowel: /æ/ is far and away more common for this. the IPA recommendation that the latter is front near-open is universally ignored. i agree that [æ] is not a usual pronunciation of ա in SEA, but then i invite you to listen to the famous song քամի փչի, to when the guy says թագուհի (within the first minute), with a noticeably fronted ա, and the other words in that section. it's pretty telling that american english /ɑ/ is far too deep for the analogous armenian sound.

anyway since the papers you sent use /ɑ/ i have no choice but to bow to the authority of convention and change it back. i will put [ɑ̈] in the phonetic representation, though.

the talk that ր involves contact with the roof of the mouth sounds crazy to my ears, because this used to be my habit, but my speech was actually corrected as a child such that i no longer do this. i was specifically told that the tongue hovers right behind the teeth and air is passed over it, which i later learned is exactly [ɹ]. so this is a hill i will die on.

but, however, since all papers talk of a *spirantized flap* (i.e., an approximant...) as a very common pronunciation of the ostensible /ɾ/, though, i will be forced to change it back. but i will also include the basically-universal spirantization in the phonetic representation in word-initial (affects only 2 words) and -final position (interesting places for lenition... but i digress).

(tangent: is there no paper talking about the word-final devoicing of certain armenian consonants?)

schwa point: i am convinced, i agree on that point. i merely wanted to express the fact that SWA's schwa is definitely higher and backer than SEA's.

dental point: you have convinced me to do the opposite with your argument... if you're saying it's the default pronunciation, then i will indicate so. idiosyncratic variation is beside the point. english coronals are well characterized as alveolar, but i know this one guy who does dentals very consistently (and he doesn't speak one of the dialects that are known to do dentals), and i also will sometimes hear a stray dental in rapid speech. this doesn't change the fact that the default pronunciation is dental, nor does it change the fact that in other languages on wiktionary, this default pronunciation is indicated.

breathy point: isn't the yerevan pronunciation essentially default, though, because armenia comprises only of yerevan? joke aside, this means that investigation in the pronunciation of բ դ ձ ջ գ by speakers of SEA from outside yerevan is necessary, investigation that does not exist. so it shall be ignored for now, until (unless?) study on that is made.

glottal point: basically the same as the previous. except i also want to clarify that i was not talking about ejectives, which i already knew about, but glottal tension, since պ տ ծ ճ կ are reported to come with some tenseness in their pronunciation, and also considering the fact that in armenia they're considered harder than փ թ ց չ ք (in contrast to what second-generation diaporans say, which is the opposite!).

also it seems that the english transcription is using a very specific analysis of syllabicity that i'm pretty sure is not accepted into consensus, which is very interesting, and that it artificially upholds the light and dark /l/, which is gone from most accents. i'm glad we hold ourselves to a higher standard on the armenian side. RagingPichu (talk) 19:47, 3 May 2024 (UTC)

Utility: if i was to see a phonetic vs phonemic distinction for a language entry, i would expect that the distinction would be about well-known systematic phonological rules that affect potentially any word in the lexicon, or a idiosyncratic pronunciation spell for some word. For example, the English examples illustrated syllabic liquids and flapping, both of which are systematic phonological rules. For the example of idiosyntratic, see the attested reduced forms of probably. As a person who uses wiktionary for research purposes (check out Wikipron if you're interested), I expect to find rational reasons like this for different levels.

/a/: in older literature for any language, people often just use the symbol <a> to mean "a low vowel" without special attention to its exact location. For Armenian, whenever I've seen <a> it's being used in a generic non-exact sense. Heck even if you look at Adjarian 1911, you'll see he uses ա <a> for a default low vowel of unspecified rounding or backness, while ա̈ is fronted and the vowel in Van is implied to be rounded. This all means that he likewise was using ա to mean default low back unrounded but implicitly. But still, we're using IPA on wiktionary, so we follow the IPA meaning of symbols such that a phonemic /a/ means front, otherwise it's just laziness.

/a/ examples: for Armenian (outside of Iran), the default pronunciation of ա is default back unrounded. But that doesn't mean you never see a fronted version in some words. For my WA in Beirut, I've never encountered /æ/ or rounded /ɒ/. But I was told by linguist colleagues in Armenia that the word ախպար can be heard with rounding /ɒχpɒɾ/ alongside standard /ɑχpɑɾ/. Such an idiosyncratic version can be set as a phonemic or phonetic version of the word: so /ɑχpɑɾ, ɒχpɒɾ/ or /ɑχpɑɾ/ and [ɑχpɑɾ, ɒχpɒɾ] with a note explaining the idiosyncraticity. For the takuhi word in that song, again, is a systematic pronunciation for that word, or an idiosyncratic way that only this guy says it, or a systematic regional aspect of some area in Armenia? A dictionary should be exact on how it's gathering its pronunciation entries.

[ɑ̈]: typical IPA is /æ/. And if you're gonna use in some phonetic representation, again, their either systematic pronunciation across the lexicon (like apple) OR an idiosyncratic but common pronunciation for a specific word (like probably)

rhotic: instead of dying on a hill, it's better to just go do a phonetic experiment :P Tabita Toparlak and Talia Tahtadjian have done past Armenian rhotic studies in case you're going down that route.

devoicing: nah the closest you get is me mentioning in a WA grammar that I'm writing (a draft online somewhere; it takes a while to write)

dental: well then if you do a dental [t] in phonetic form because systematic, then a question is why have non-dental /t/ in phonemic form? One would expect that phonemic form is more careful speech in some way, but if the /t/ is always dental, then I'd expect a different symbol or just an explanation in the key. Oftentimes in the linguistic world that I see, adding dental diacritics are too tiring so a default /t/ with a footnote often suffices.

English liquid: nah syllabic sonorants are a common thing in English linguistics, to the point that we even teach it to undergrads. https://www.google.com/search?client=firefox-b-1-d&q=syllablic+liquid+english Hovsepig (talk) 17:32, 4 May 2024 (UTC)

regarding utility: i would expect the same from a linguistic resource, which wiktionary arguably is indeed, but it's not geared toward linguists, meaning that what's extraneous to you and i may help someone else. basically, the extra information will not hurt anyone. if the diacritics are undesirable in wikipron, they can simply be cleaned with ctrl+H during pre-processing of the data, which would be necessary anyway for normalization.

regarding /a/: then someone should edit the american english /æ/ to be /a/ instead, for this vowel is often rather low and not near-low. but no one will. basically, what i usually see is transcription depending on how many low vowels a language has: one low vowel means /a/, two means /æ ɑ(ɒ)/, and three means /æ a ɑ(ɒ)/ (or sometimes /ɛ a ɔ/...), with the exception for the first group being languages like turkish where the low vowel explicitly behaves back in the harmony system. armenian doesn't have evidence for /a/ phonologically behaving like a back vowel, as far as the analyses i've read at least.

regarding [ɑ̈]: this symbol does not mean a fronted ɑ, but rather a centralized ɑ, according to the IPA. an illustrative example is high vowels: from back to front we have [u ü ʉ ÿ y]. likewise from back to front for low vowels we have [ɑ ɑ̈ a æ̈ æ]. we both know that phonemically these distinctions are never useful, because such granular distinctions are thought to be impossible. that they exist in the IPA at all, though, show that they are useful in a phonetic context to represent the most neutral pronunciation, since, although not contrastive, the difference is still audible.

the pronunciation as central-back is systematic in neutral, educated speech. pronunciations otherwise are due to other factors; cf. the pronunciation s[a]rry (for ordinary s[ɑ]rry) in american english expressing insincerity.

tangentially, [ա̈] is ugly, and [ՠ] is better... i don't understand what bewitched linguists to abandon it. i can understand [ո̈ւ օ̈], though.

i expect that [ա̊խպա̊ր] comes from a certain emotive style of speech that deserves study. this pronunciation is not neutral. you will hear it when people complain, for example, as in [ա̊խ ինչ ցա̊վում ա̊ ձէռքըս]. i associate it with women, but this is just anecdotal, and it requires investigation.

թագուհի seems to be fronted because of the singing style. the point i was trying to make is that this pronunciation is still perfectly understandable to an armenian, meaning that frontness-backness is not contrastive.

rhotic: i will do both.

devoicing: unfortunate... that seems like another experiment waiting to be done.

dental: the same laziness that makes people write /r/ or /u/ for american english, a laziness that is actually sanctioned by the IPA, which is why it's still common. we can show that phonemic form is not more careful speech by considering american english æ-tensing, or american raising (canadian raising without /aʊ/ affected), or the backing of /u/ before coda /l/, which are all mandatory no matter what (no matter how careful the speech, pan and pad, write and ride, and rude and rule do not have the same vowel phonetically for speakers where these rules are operative). we may also consider the typical transcription for romance languages, which all have dentals for their /t/, but are never transcribed so in phonemic transcription, for exactly the reason you gave, but that wiktionary for spanish (and turkish, inconsistently...) represents them as dental in phonetic transcription.

tangentially, the lack of separate symbols for dental and alveolar stops in the IPA is beyond ridiculous, since the distinction is known in many languages from many places. this violates the IPA principles. the click consonants fare better in this respect, believe it or not.

and, i was specifically referring to /ˈlæt.əɹ/ with the /t/ closing the syllable. sorry for being unclear. i do not dispute that the /l/ in apple is syllabic: that's definitely non-controversial.

by the way, i will make the edits i said i would make probably in two or four days. i haven't forgotten about them. RagingPichu (talk) 09:51, 5 May 2024 (UTC)

@RagingPichu also revert the lack of aspiration in WA like կարագ. Whether voiceless stops are aspirated varies by geographic region. But the convention is to follow Istanbul pronunciations for aspiration marking in dictionaries. Hovsepig (talk) 11:35, 11 May 2024 (UTC)

@RagingPichu: Why did you split the phonemic and phonetic transcriptions into two modules? You should merge them into one and the template code should be rewritten to support it. — Fenakhay ^{(حيطي · مساهماتي)} 00:13, 12 May 2024 (UTC)

i have actually been thinking of how to merge them, since they obviously share a lot of overlap. the way i did it now was the quick way while i think of a more elaborate method to handle it, and then deprecate Module:hy-phonemes. i was planning to do this after i finish creating a conjugation module that can automatically detect which type of verb it's dealing with. RagingPichu (talk) 00:19, 12 May 2024 (UTC)

I have reverted the recent edits, not because of the details of the phonetic transcription (I don't know enough about Armenian to have an opinion) but because it was showing the alternative pronunciations thus: {{IPA|hy|/.../ [...]}}, which has completely overwhelmed both Category:IPA pronunciations with invalid representation marks and Category:IPA pronunciations with invalid IPA characters. If you want to show multiple pronunciations within a single {{IPA}} tag, you have to put them in separate parameters thus: {{IPA|hy|/.../|[...]}}. —Mahāgaja · talk 19:40, 11 May 2024 (UTC)

thanks, i fixed it (before @Fenakhay reverted it). this brings up another issue, though: @Vahagn Petrosyan added in the allowed set of armenian phonemes into Module:IPA/data nearly 10 years ago, which is (was) generating red marks for words like and Էօժէնի (Ēōžēni) since 'e' and 'œ' aren't recognized in the list of armenian phonemes, a list which should be updated (to include potential dialectal data as well, for the future). RagingPichu (talk) 00:17, 12 May 2024 (UTC)

@RagingPichu: I have updated Module:IPA/data for 'hy'. Let me know if any phonemes are missing. Vahag (talk) 08:09, 12 May 2024 (UTC)

@RagingPichu, why is your code generating Eastern Armenian /kʰɑˈtʰɑ/ for գաթա (gatʻa), and not /ɡɑˈtʰɑ/? --Vahag (talk) 12:19, 25 May 2024 (UTC)

very bizarre... i'll seek other examples to see what the pattern is. it's also generating /ɡɑˈkʰɑtʰ/, [ɡɑɡɑ́tʰ] for գագաթ (gagatʻ). i'll update when i fix it. RagingPichu (talk) 21:34, 25 May 2024 (UTC)

i fixed it. the problem ultimately lay in the regex capturing CVCʰ as an aspirating environment for the voiced stops, since it resembles the environemnt to trigger rule i made designed to capture aspirated affricates. i blame the IPA's silly convention :). delaying the conversion to IPA and processing the aspiration with armenian letters seems to have fixed the problem, and made the code shorter, too. RagingPichu (talk) 22:01, 25 May 2024 (UTC)

badly written code

(Notifying Vahagn Petrosyan, Agajania, Hovsepig, Pompyxmori): @RagingPichu There's some badly written code here that needs cleanup. The choice to split Module:hy-pronunciation and Module:hy-phonemes was clearly wrong. In addition, we have both {{hy-pron}} and {{hy-IPA}} (why? what's the difference?), and both of them have a mixture of template and module code, when everything should be in (one single) module. I am going to try and clean some of this up, at least move the template code into the module; but User:RagingPichu you will need to re-merge the modules. Benwing2 (talk) 02:03, 21 May 2024 (UTC)

@Benwing2: it looks like @Fenakhay cleaned up after us (thank you). {{hy-IPA}} generates the pronunciation, and {{hy-pron}} slaps on dialect tags on it. If this functionality can be merged into one template, that would be great. Vahag (talk) 14:19, 21 May 2024 (UTC)