Template talk:gl-pr

New Galician pronunciation template

@Stríðsdrengur, Rodrigo5260, Froaringus: The Galician pronunciation template is mostly done, so this is probably a better place to discuss it. My sources have been Gramática da lingua galega (I). Fonética e fonoloxía by Freixeiro Mato and the Dicionario de pronuncia galega. The actual usage of the template is documented here, so instead I'll try to summarize some (potential) issues with it:

• Compulsory marking of the quality of /e~ɛ o~ɔ/ in pre-tonic and tonic syllables. With some exceptions, the template can't predict whether a vowel written with <e o> is mid-open or mid-closed, so the user must use an appropriate respelling. If the template runs into an undetermined vowel, it will throw an error (the Portuguese template does something similar).
• Word-final unstressed vowels. The template shows them as phonemic /ɐ ɪ ʊ/. In some sources they are transcribed with the symbols e̝, o̝, a̝ instead. While this notation makes the transcription more similar to the spelling, I think the transcription without diacritics is less cluttered.
• Inclusion of gheada/seseo. The template will generate a pronunciation with gheada and/or seseo whenever it encounters /θ/ or /ɡ/. These alternative pronunciations are hidden by default, so unless the user clicks on "Show", they will remain invisible.
• Phonetic transcription. For reference, I've tried to keep it at about the same level of narrowness as in the Spanish template.
  • Nasalization of vowels. The template will nasalize a vowel if followed by a nasal consonant in the same syllable or if preceded by /ŋ/.
  • Allophones of /a/. Depending on the context, the vowel will be shown as [a ɑ a̠]. I have two doubts regarding these allophones. First, is word-final /a~ɐ/ after a velar consonant pronounced [ɐ] or [ɑ]? Secondly, should we treat long vowels such as the one in ra (“frog”) as phonemically long /aː/ [aː] or as a sequence of two vowels /aa/ [aː]?
  • Allophones of /s θ/. They are voiced to [z θ̬] before voiced consonants. In general, /s/ (and its voiced allophone) will be marked as apico-alveolar, except before palatal consonants where it will be shown as palatalized [sʲ].
  • Assimilation of /n l/. They will be shown as [n ɱ n̪ nʲ l l̪ lʲ ɫ] depending on their position. Word-finally and in other environments the nasal consonant is shown as phonemic /ŋ/.
  • Lenition of /b d ɡ/. They will be shown as [β̞ ð̞ ɣ̞] where appropriate.

I've never edited Galician, so I'll be happy to hear your comments and suggestions before moving it to the mainspace. Santi2222 (talk) 13:10, 4 March 2024 (UTC)

I am very grateful and happy for your work! I think the module is already good enough, and if we have to change something we will let you know or we will do it ourselves if it is not too big. Thanks again for your work! Stríðsdrengur (talk) 13:12, 4 March 2024 (UTC)

Excellent work! Main issue I had in my mind is that a stressed vowel can be either open or closed in Western / Eastern Galician (the later usually agrees with the expected etymological results, the former tends to agree with Portuguese) but since the module doesn't assume its values, there's no problem.

• Final or after tonic /a~ɐ/ after a velar consonant should be pronounced [ɐ].
• The long /a/ is due mostly to the loss of intervocalic l or n, so they are usually hard to predict, and speakers tend to reduce them; I would, generally, leave them out.

Additional considerations:

• Fernando Martínez Gil, "Galician", in Manual of Romance Phonetics and Phonology (via the Wikipedia library!)
• https://www.sciencedirect.com/science/article/abs/pii/S0167639308000228

Froaringus (talk) 14:07, 4 March 2024 (UTC)

Thank you @Froaringus! I've made two small changes to the module that powers the template: final unstressed /a~ɐ/ will always be [ɐ] regardless of the preceding consonant and vowels surrounded by nasal consonants will be nasalized. In fact, it seems like nasalization triggered by a syllable-final nasal is weaker than the one triggered by surrounding nasals. Should we remove it from the transcriptions? One of the sources you provided states that "[the] process is highly variable" and they omit it from their transcriptions. Also, the template won't support long vowels for now, but it would be easy to add them if needed.Santi2222 (talk) 15:06, 4 March 2024 (UTC)

I would leave nasalization triggered in between nasals, because among older speakers is very very notable, and prescind of its notation in other contexts... In general, Galician sounds much more nasal than Spanish. Anyway this is very speaker-specific! Froaringus (talk) 16:47, 4 March 2024 (UTC)

I've removed nasalization except when triggered by surrounding vowels. The template is ready so I'll move it to the mainspace, but of course feel free to ping me if you find any bugs. Santi2222 (talk) 16:42, 5 March 2024 (UTC)

Corrections

Hi, @Santi2222. Could you correct the module/template so that the result of

{{gl-pr|brètema}}

is

IPA(key): /ˈbɾɛtema/ [ˈbɾɛ.t̪ɪ.mɐ]
Rhymes: -ɛtɪmɐ

That is: in phonemic transcription final /a/, and in phonemic [ɪ ɐ ʊ] for any vowel in any syllable following the stressed syllable?

Many, many thanks! Froaringus (talk) 13:51, 7 March 2024 (UTC)

So, /e o a/ should be reduced to [ɪ ʊ ɐ] in post-tonic syllables (not just word-final syllables) if I'm understanding correctly?Santi2222 (talk) 11:16, 8 March 2024 (UTC)

Yeah, that's what he meant. Rodrigo5260 (talk) 16:50, 8 March 2024 (UTC)

OK, I'll get to it. I suppose that /i u/ shouldn't be reduced, right? According to the sources I've consulted there are five contrasting vowels in post-tonic syllables. Santi2222 (talk) 18:48, 12 March 2024 (UTC)

I've just changed the module, vowel reduction will be shown in phonetic transcriptions only. Santi2222 (talk) 21:17, 21 March 2024 (UTC)

If it helps

Hi everyone. A while ago I wrote out some phonetic transcriptions and descriptions taken from the book Manual of Romance phonetics and phonology, which has a chapter on Galician (page 882 onwards). What I collected can be found here. Perhaps it can be useful. Nicodene (talk) 00:03, 22 March 2024 (UTC)

The "almuína" case in the documentation

First of all congrats for making this template, @Santi2222 — it's a very pleasant surprise for sure!

Still, I have a bit of feedback: I don't think a syllable separator should be needed for "almuína"; the accent on the I, shown by the spelling, already specifies that it's /mu'i/ and not /'muj/. Would it be possible to tweak the module so that acute accents become able to do syllable breaking? The Portuguese and Spanish modules both have it, so perhaps one could look at those as reference for how to code 'em... MedK1 (talk) 13:46, 14 July 2024 (UTC)

Hello @MedK1, thanks for the input. The handling of that particular sequence is intentional, the issue here is that the sequence <uí> can stand for both /wi/ (as in acuícola) and /uˈi/ (as in almuíña), so a manual respelling will always be necessary for some words. Personally I feel like it is more natural to use a dot to mark a hiatus while keeping the rest of the spelling rather than having to rewrite <u> as <w>, but of course this could be open to discussion . Santi2222 (talk) 14:35, 14 July 2024 (UTC)

That makes sense. Good thinking! I actually believe a solution where neither word needs to be respelled can be achieved, however: Currently, the situation seems to be handled by the excerpt of code after "we will use /wi/ after velar sounds and /uj/ elsewhere", which is a really good solution. What I'd propose here, though, is to add some sort of check for the non-velar consonant route for, if the I is accented, then it makes the change to /uˈi/, instead of /ˈuj/.

I really do think they're not mutually exclusive; the main 'problem' can be demonstrated by this:

{{gl-pr|almuína}}

• IPA^(key): /alˈmujna/ [ɑlˈmuj.nɐ]
• Rhymes: -ujna

This prompt generated an impossible pronunciation; there is no case in which a person could read "almuína" and produce /alˈmujna/. The code distinguishes between /wi/ vs /uj/, and in terms of "words where the I here is accented", the former covers some of those cases while the latter covers no such cases. If it checked for an accent and in cases with an accent distinguished only between /wi/ and /uˈi/ (instead of /wi/ and /uj/), that'd be — and I hope you'll agree — an optimization of the code with no downsides; it should ideally completely eliminate one case where you'd normally need to do rewriting. (A case like "luita" would be unaffected since the I is not accented) MedK1 (talk) 14:56, 14 July 2024 (UTC)