Module talk:grc-pronunciation/Archive 2

The discussion here, particularly regarding moving the "expand" bit closer to the rest of the text. — LlywelynII 04:15, 31 July 2015 (UTC)Reply

@LlywelynII: I agree that should be done, but I tried and couldn't figure out how to do it. I tried setting the width of the largest div <div>, but then the transcriptions end up wrapping, if the word being transcribed is very long. There doesn't seem to be a way for the div to adapt to the size of the stuff inside it. Or maybe I just haven't considered all the options. Maybe the only way would be to make the module count the number of characters that will be displaying in the collapsed and the full transcriptions, and then set the width to that number of em. But it might be complicated to count the number of displayed characters when there's all this HTML being added to it in the output. — Eru·tuon 03:58, 19 October 2016 (UTC)Reply

Actually, some more tinkering and I figured something out. Hopefully it displays as well for other people as it does for me. — Eru·tuon 04:25, 19 October 2016 (UTC)Reply

I'm not sure if these things are wrong, but comparing some of this template's results with some of {{grc-ipa-rows}}'s results makes me wonder about some things. First of all, consider when did the voicing of π τ κ after a nasal take place? {{grc-ipa-rows}} shows it in 15th c. Constantinopolitan, but {{grc-IPA}} doesn't:

Which is right?

Second, the treatment of β δ γ after a nasal is not consistent in {{grc-IPA}}:

Did the voiced stops really first become fricatives and then become stops again (with the exception of unpalatalized gamma, which apparently remained a fricative)? Did the nasal consonant really disappear before /b/ and /d/ but remain before /ɟ/ and /ɣ/? —Aɴɢʀ (talk) 20:00, 28 January 2016 (UTC)Reply

@Angr: This is a tough question, the pronunciation of stops after nasals. I'm not sure how one would go about determining it. But I suppose it would be best to simply treat all the sequences the same, phonologically speaking, in each period of Greek, unless there is evidence that they were different. — Eru·tuon 22:07, 12 March 2017 (UTC)Reply

No, voiced plosives remain voiced after homorganic nasal consonants through the end of the Ancient period in 1453. And yes, this seems to imply that, at some point, voiced and voiceless plosives after homorganic nasals did indeed merge, as most cases of γγ have merged with γκ in modern Demotic. And if I recall correctly, the lenition of plosives after homorganic nasals was a rather artificial development that emerged with Katharevousa and rubbed off onto Katharevousa-influenced modern Demotic Greek. Of course, this is just what I remember off-hand, but it does reflect how I wrote these templates before they were converted to Lua. I'd edit them myself now, but I never learnt Lua and I wouldn't really know what I'm doing. - Gilgamesh~enwiki (talk) 04:42, 24 October 2017 (UTC)Reply

@Gilgamesh~enwiki: Thanks, do you happen to remember what sources you used? Also, @Barytonesis, Erytuon. —JohnC5 09:05, 24 October 2017 (UTC)Reply

@Erutuon --Barytonesis (talk) 09:12, 24 October 2017 (UTC)Reply

Oh, just what's already been provided in Wikipedia's articles for a long time now. I tried to follow the details meticulously, providing for all the conservatively regular sound laws as documented, which informed how I wrote Template:grc-ipa-rows and its dependent token-based Template:grc-ipa-cla (5th BC Attic relying on w:Ancient Greek phonology), Template:grc-ipa-koi (1st BC Egyptian relying on w:Koine Greek phonology), Template:grc-ipa-koi2 (4th BC Koine relying on the same article), Template:grc-ipa-byz (10th BC Byzantine relying on w:Medieval Greek) and Template:grc-ipa-byz2 (15th BC Constantinopolitan relying on the same article). Though I'm not saying my choices of token logic certainly shouldn't be reviewed as needed, I based it on the sound laws documented, for example:

• For the sequence εῦφ, indicated by the token "eeu^ph":
  • {{grc-ipa-cla|eeu^ph}} → ɛ͜ɛ́͜ʊpʰ (because the vowel is metrically at least two morae long and the tone falls on the penultimate vowel mora instead of the last vowel mora)
  • {{grc-ipa-koi|eeu^ph}} → éːʍpʰ (because the off-glide became a consonant, devoiced before voiceless consonants)
  • {{grc-ipa-koi2|eeu^ph}} → íɸ̚ɸ (because the off-glide and the following consonant merged into a long consonant)
  • {{grc-ipa-byz|eeu^ph}} → íf̚f
  • {{grc-ipa-byz2|eeu^ph}} → íf (because geminated consonants had lost their gemination in Constantinopolitan at the time of the fall of Constantinople)
• The 2007 version of my token templates (broken in my 2012 template restructuring which I've at least partially bugfixed just now) had tokens like ns (after long vowels) and nns (after short vowels), where a nasal before a sonorant was unstable in Classical Greek and would regularly disappear and lengthen the preceding vowel, which was actually anticipated in names like Τίρῠνς (the ί is actually accented long ῑ), which was actually pronounced *Τίρῡς with two long vowels, otherwise Classical Attic tonal laws would have forced it to be *Τῖρῠνς. However, this elision of nasals was no longer necessarily productive in the learnt pronunciation of the Hellenistic period when many non-Greek peoples learnt Greek, and so the nasal returns in pronunciation, though disappears again in Constantinopolitan when nasals regularly disappeared before all fricatives. So {{grc-ipa-rows|t|ii'|r|u|nns}} became:
  (deprecated template usage)

• (5th BC Attic) IPA^(key): /ti͜íryːs/

• (1st BC Egyptian) IPA^(key): /tíːryns/

• (4th AD Koine) IPA^(key): /tíryns/

• (10th AD Byzantine) IPA^(key): /tírỹs/

• (15th AD Constantinopolitan) IPA^(key): /tíɾis/

• Also regular sound laws for voiceless plosive-plosive and voiceless fricative-fricative consonant clusters becoming fricative-plosive sequences (or plosive-fricative sequences if the second consonant was /s/), which during even learnt Medieval Greek was a function of basic language phonotactics of the time (and also allowed late Koine ξ ψ to become */ks ps/ instead of */xs fs/ from Attic and earlier Koine /kʰs pʰs/). Language purism in a medieval Byzantine Empire was largely a matter of maintaining prescribed written norms, and how it was spoken was not as such a priority as the language phonotactics kept evolving. It wasn't really until the much later Katharevousa period with a more centralized Greek state education system that there was real success in changing educated speaker phonotactics to allow sequences like plosive-plosive, fricative-fricative, nasal-fricative, etc.

I took care to write the documentation in Template:grc-ipa-rows/documentation to assist future editors (for both embedding the rows template and maintaining the template and tokens themselves), though this was before Lua-based template scripting really became a thing. I know template maintenance is necessarily a more technical topic with fewer takers than the editing of articles themselves, because one has to be both sufficiently educated in the linguistics topic as well as in the more technical aspects of template embedding syntax, logic and (now also, obviously) Lua scripting. When I wrote the documentation and tried to make the templates as maintainable as I could, it didn't occur to me that this new programmatic scripting approach would see someone rewriting the template in Lua with a haphazard replication of my token logic. - Gilgamesh~enwiki (talk) 17:17, 24 October 2017 (UTC)Reply

I've made voiced stops remain stops after nasals according to what @Gilgamesh~enwiki said about them above. There is still the inconsistency in which the nasal before the stop vanishes in the /mb nd/ sequences. — Eru·tuon 23:35, 2 November 2017 (UTC)Reply

Now γγ is transcribed as /ɡ, ɟ/ (diff1), and μπ, ντ, γκ as /b, d, ɡ, ɟ/, in byz2 (diff2). — Eru·tuon 21:49, 18 December 2017 (UTC)Reply

A problem I just noticed: the template doesn't accept combining acutes. λᾰνθᾰ́νω (lanthánō) currently renders without pitch accent or stress marks: /lan.tʰa.nɔː/ → /lan.θa.no/ → /lan.θa.no/ invalid IPA characters (/→//→/). — Eru·tuon 06:16, 21 February 2016 (UTC)Reply

Yeah, you have to do it the other way round: ά plus a combining breve. It would be nice if it recognized both variations, but it doesn't. —Aɴɢʀ (talk) 10:20, 21 February 2016 (UTC)Reply

As I recall, W. S. Allen says γ before μ was assimilated and pronounced as a nasal: πρᾶγμα . γ before ν might as well, but evidence is less clear. The occurrence of γν (but not γμ) in syllable onsets might suggest γ is a stop there, though perhaps assimilation happened between vowels. Different situation from Latin, where g in gn was clearly assimilated.

Just a note. Another item on the list, I'm sure (thinking of ). — Eru·tuon 05:06, 27 February 2016 (UTC)Reply

Hi @Erutuon, can you use these templates in the pron please? Thanks in advance! — AWESOME meeos ！ * (не нажима́йте сюда́ ) 08:47, 15 March 2017 (UTC)Reply

The text that includes "BCE", etc. is already a link itself though. —suzukaze (t・c) 08:50, 15 March 2017 (UTC)Reply

When the templates are used, users can decide for themselves (thanks to Per-browser preferences) whether to have it displayed as "BC" and "AD" or "BCE" and "CE". —Aɴɢʀ (talk) 14:01, 15 March 2017 (UTC)Reply

This is not handled by mod:grc-pronunciation but instead by mod:a/data. It's not immediately clear to me the best way to implement this change. —JohnC5 14:59, 15 March 2017 (UTC)Reply

Finally done. I created an alternative form of this date label, because the original one would not work in the template, and a separate procedure in the JavaScript gadget to handle it. — Eru·tuon 01:48, 1 December 2017 (UTC)Reply

Thank you so much. We appreciate. sarri.greek (talk) 04:33, 2 December 2017 (UTC)Reply

I believe they were lost earlier, see, for example, this papyrus from the 6th century with κομέντο(ν) for κομμέντω(ν). Does anyone have more data on the topic? Guldrelokk (talk) 12:40, 6 April 2018 (UTC)Reply

• 1. Are these correct? nst now is ns.t Would n.st be wrong? e.g.

Same with str (s.tr instead of .str) Oh just noticed: is the k.s not ks?

• 2. Is it possible to isolate periods? Or at least, stop the later ones appear? e.g. I need only cla for e.g. ἔχεν

Thank you, sarri.greek (talk) 04:28, 18 August 2018 (UTC)Reply

@Sarri.greek: At the moment, you can only choose which period to start with. My reasoning in support of this is that even if the word wasn't used, Greeks of later ages still read it out loud in their pronunciation, like the YouTube video I once watched of Homer recited or sung with Modern Greek pronunciation. So, even Homeric words have a Koine or Medieval pronunciation.

We haven't made an attempt to divide syllables correctly yet. I don't even know how the module does it. I would want to base it on syllable weight: if a syllable was heavy (long) and didn't have a long vowel, it had to end in a consonant. For instance, regarding the ks example, ᾰ̓́ξῐος (áxios) would have been heavy–light–heavy in Homer, and the weight of the first syllable means that the ks cluster has to be split between syllables.

I guess that's different from the practice of letting a word-internal consonant cluster be a syllable onset if it is also found at the beginning of a word. Under that practice, since ks can appear at the beginning of a word (like ξένος), it shouldn't be split between syllables: /á.ksi.os/, light–light–light. But then the IPA transcription would disagree with actual syllable weight. So /ák.si.os/ is better, and that practice can't be applied to periods of Greek that had quantitative meter.

Three-consonant clusters too have to donate at least one consonant to the previous syllable. For instance, ᾰ̓́στρα (ástra) is long–short, so /ás.tra/ or maybe /ást.ra/, but not /á.stra/. There are clusters that are variable though, which is part of why we haven't automated it yet. — Eru·tuon 08:39, 18 August 2018 (UTC)Reply

Anyway, in response to your first question, I agree with you that n.st looks better than ns.t. I might work on fixing that. — Eru·tuon 08:43, 18 August 2018 (UTC)Reply

Thank you @Erutuon. I have no opinion on phonology. And I admire immensely your work here. From what you are telling me, my impression is

• 1. The 5 periods should be dealt with in isolation. Non-prosody periods, very different, as you describe. 5th period has only slight difference with the 6th period: modern. (y and some double consonants...). Ξ, Ψ split in 10th century? to k‑s and p‑s? (compounds ἔκστασις ok) But I eat an op‑sarion (I eat a fish ὀψάριον)? They are double‑consonants by now, not clusters?
• 2. Manual intervention should be possible -for authorized editors; like you- For minor-corrections. E.g. Hard to believe that in 15 century, or any non-prosody-period, they were pronouncing compound words like ἀ‑διάσ‑τροφος instead of ἀ‑διά‑στροφος (from στροφή and not τροφή). I have no opinion for consonant clusters within stems.

You are doing such an excellent work! I just wrote here to give my feedback, but it is a speaker's feedback, not a phonologist's. Eru! If we go back to 5thBCE Athens, everyone will understand YOU and not me. But if we go to Alexandria, or Constantinople, of 3rdCE I think they will understand me a bit more?! sarri.greek (talk) 14:31, 18 August 2018 (UTC)Reply

I noticed that double consonants are shown in Egyptian, Koine, and Byzantine transcriptions as geminate consonants broken across syllable boundaries (e.g.: ;Ἀπολλῡ́ων; /a.polˈly.on/). However, I could not find any information on the double consonants' actual gemination—as in having a pressure contour of two distinct peaks, one for either syllable—and I have read it proposed elsewhere that the consonants were probably just long. Should this be reflected in how the template transcribes these consonants, that /…C₁.C₁…/ would instead become /….ːC₁…/ (e.g.: */a.po.ˈːly.on/)? BlueCaper (talk) 20:55, 10 October 2018 (UTC)Reply

@BlueCaper: Writing a geminate consonant with two symbols isn't meant to imply anything about pressure contours. I think we went with it for Ancient Greek because it allows the transcription to clearly show syllable weight. Ἀπολλῡ́ων (Apollū́ōn) /a.pol.ly̌ː.ɔːn/ is light–heavy–heavy–heavy, and /pol/ is clearly a heavy syllable. The transcription /a.po.lːy̌ː.ɔːn/ suggests that the second syllable is light. (The length mark goes after the consonant.) So if the length symbol was used, the transcription would not show syllable weight before a double consonant.

Syllable weight is a significant phonological feature that determines the form that the comparative and superlative suffixes take when added to first- and second-declension adjectives: -ότερος (-óteros) and -ότατος (-ótatos) after a heavy syllable, or -ώτερος (-ṓteros) and -ώτατος (-ṓtatos) after a light one. (Granted, I couldn't find an adjective with a comparative and superlative that has a stem ending in a double consonant.) It's also the basis for quantitative meter. So it's helpful to show it in the transcriptions. — Eru·tuon 22:08, 10 October 2018 (UTC)Reply