Module talk:de-IPA

Bad idea?
I think German spelling is too irregular to be able to deduce the pronunciation reliably. —CodeCat 17:29, 15 November 2014 (UTC)

Variants
The inclusion of aspirations was decided against in the past, presenting the fortis-lenis constrast as one of voicing here. But German linguistics as far as I know it - which is limited but more than nothing - treats voicing as facultative, and I read more than one author who assumes /d, b, g/ to be voiceless per default. Hence my move to include aspiration to mark fortis consonants in non-southern German German. The reason given for excluding aspiration was that not all areas in Germany aspirate, but in the light of the recent discussion in the Beer Parlour, I would say that the lack of aspiration is outside of the scope of the standard variant this module is for now meant to portray. As of now, this is mainly an information for future reference, however. Actually implementing aspiration rules into the module, should that be decided, is certainly nothing for the first step. Korn &#91;kʰũːɘ̃n&#93; (talk) 11:20, 11 October 2016 (UTC)
 * It seems to me that the situation with voicing vs. fortis-lenis is not so different in German as in English; in both languages, the voiced stops are sometimes pronounced unvoiced or partly unvoiced, and the unvoiced stops are often aspirated. For this reason it makes sense to me to use a voicing contrast in IPA, since English speakers (who are the target audience here) will know what to do instinctively. Using p vs pʰ would seem more confusing to me than b vs p. But if most dictionaries do it the other way, I could be persuaded to think differently. Benwing2 (talk) 14:43, 11 October 2016 (UTC)
 * You got a fair point, so I won't try to persuade you. Plain dictionaries should normally portray a voice/unvoiced contrast; the whole fortis-lenis matter is only paid attention to in detailed treatments. I might bring this back up should I tackle the south, which doesn't have voiced obstruents. Korn &#91;kʰũːɘ̃n&#93; (talk) 19:07, 11 October 2016 (UTC)

indicate glottal stop word-internally?
I wonder if we shouldn't indicate a glottal stop word-internally. I remember a discussion of this that pointed out that Himmelsau "Heaven's eye" has a glottal stop in it, and without it it would mean "heavenly sow". Aufenthalt is written [ˈaʊ̯fʔɛnthalt] in dewikt, where the glottal stop is not predictable unless you know to analyze the word into auf- + enthalt. Benwing2 (talk) 07:05, 12 October 2016 (UTC)
 * The glottal stop is merely an allophone of the zero onset of a syllable. In your example, it's the audible representation of a syllable break. Himmelsau being /himel.sau/ [hɪməl(d)zaʊ] or /himels.au/ [hɪməltsʔaʊ]. But the stop does not actually contrast with a plain unstopped hiat [hɪməltsaʊ]. I use [ʔ] consistently and phonemically (because zero onset is an allophone of /h/ for me) but drop it inside of words in rapid speech. And I think, but don't know, it's not used somewhere south. So I'd put it somewhere in the area of aspiration: It's a thing, but it's only one way to express a bigger phænomenon clearly, and it's not universal. So if you're not indicating syllables, you might want to include the stop, though I think its presence is predictable, for clarity.
 * Also [ˈaʊ̯fʔɛnthalt] sounds literally wrong to me. I wanted to call it hypercorrect, but that would imply I would have seen the possibility to parse it as /auf.ɛnt.halt/, which isn't the case, I would have never gotten even the idea. I have never heard anyone not pronounce it /au.fən.talt/ (= [aʊ̯fn̩ˈtʰalt] in local normal speech, [ˈaʊ̯fɪ̽nˌthalt] in careful speech). Your version might be Swiss or Austrian. That said, if you would have /auf.ɛnthalt/, the stop would be predictable since, in dialects which use [ʔ], syllables cannot begin with a vowel. Korn &#91;kʰũːɘ̃n&#93; (talk) 09:14, 12 October 2016 (UTC)
 * ps.: /ˈaufˌɛnt.halt/ is the pronunciation used on duden.de, but it sounds as wrong to my ears as I expected, since secondary stress is on /halt/ here. Korn &#91;kʰũːɘ̃n&#93; (talk) 09:21, 12 October 2016 (UTC)

"stag" in sequences
, under the sequences for "s", why is "stag" separate? What example is this meant to show? — JohnC5 14:27, 13 October 2016
 * Romanic -gn- violates German rules, so for example ⟨stagnieren⟩ is [ʃtagniːrən] with unclear syllable breaks, instead of *[ʃtak.niːrən] or *[ʃtaχ.niːrən]. Korn &#91;kʰũːɘ̃n&#93; (talk) 14:44, 13 October 2016 (UTC)
 * But this doesn't solve that problem. This instead leads to that ⟨st⟩ becoming [st], which is wrong. There should be a rule for ⟨gn⟩, not for ⟨stag⟩. — JohnC5 14:55, 13 October 2016 (UTC)
 * I'm guessing stag was there because of the word Reichstag, but it almost certainly shouldn't be there. As for gn, there are other words where it's pronounced /gn/, e.g. regnen, Wagner, so maybe it should be rendered as such by default. Benwing2 (talk) 17:59, 13 October 2016 (UTC)
 * The Germanic ones are pronounced with /g.n/ [χn~çn] in my native speech, i.e. with final obstruent devoicing. I don't know how other regions say it, but if it's doable, dropping in there a provision for splitting variants of that combination into Germanic/Romanic, that might not be the worst idea. Korn &#91;kʰũːɘ̃n&#93; (talk) 18:31, 13 October 2016 (UTC)
 * This should be doable; we will already be splitting v on Germanic/Romance. Benwing2 (talk) 22:17, 13 October 2016 (UTC)

Should this work continue?

 * It's a shame this development has stopped and I disagree it's too irregular if one variety is chosen - Duden could possibly be chosen as default. There is nothing terribly difficult about the standard German phonology (at least one variety), which can't be handled and you have done much more complicated ones. I encourage you to add it to your list. It would be great to have a module. I am quite familiar with the German phonology but we have better-skilled native speakers who probably will join the effort. Anatoli T. (обсудить/вклад) 23:10, 20 December 2022 (UTC)
 * Module:User:Benwing2/de-pron/testcases/misc, Module:User:Benwing2/de-pron/testcases/prefixes, Module:User:Benwing2/de-pron/testcases/suffixes. A nearly complete implementation of this module exists in Module:User:Benwing2/de-pron, but I haven't had the chance lately to do the final touches. A couple of months ago I pinged User:Fytcha about this, who responded at Module talk:User:Benwing2/de-pron/testcases/misc. I would have probably despaired of figuring out all the edge cases; this is already the most complex pronunciation module I've written, running at around 2500 lines of code, and will certainly grow. But I managed to get this done for Portuguese in Module:pt-pronunc (which also has a ton of edge cases and has additional complexities due to having no single standard variety), so maybe it's possible and can be done incrementally, like was done for the Portuguese module. Benwing2 (talk) 02:30, 21 December 2022 (UTC)
 * Amazing. I think it's especially complex because of loanwords, it would probably be much less complex with only native German words.
 * I shouldn't make assumptions about the complexity of modules. Thank you for your efforts and good luck! Anatoli T. (обсудить/вклад) 03:24, 21 December 2022 (UTC)
 * If it is of help to you, I can continue to chip away at the false-negatives in your unit tests once I have more time to work on Wiktionary again.
 * On a related note, seeing that we already have IPA data for the majority of German simplices, another idea I've had was to develop an automatic IPA and hyphenation template for German compounds and non-lemmas that draws on the IPA and hyphenation data found in the entries of their constituents (in the case of compounds) or lemmas (in the case of non-lemmas). For instance, the IPA and hyphenation data in could be automatically generated from the data found in  and  and similarly,  could be generated using the data in . The editors over at de.wikt have already implemented this idea for non-lemmas as far as I know. This is a significantly easier task that also allows a more robust solution while still covering (what I believe to be) the overwhelming majority of entries that currently lack IPA and hyphenation data. &mdash; Fytcha〈 T | L | C 〉 11:10, 31 December 2022 (UTC)