Module talk:User:Benwing2/de-pron/testcases/misc

Sandbox pronunciation module
You might be interested in this sandbox module, which I created five months ago or so but never completely finished. It implements automatic pronunciation generation for German. You can see the testcases here as well as Module:User:Benwing2/de-pron/testcases/prefixes and Module:User:Benwing2/de-pron/testcases/suffixes. If you think this is useful, I can look into finishing it. It's pretty much useful as-is but has some edge cases that maybe aren't quite handled right, as can be seen in the failing testcases. Benwing2 (talk) 04:27, 30 October 2022 (UTC)


 * Wow, I wouldn't have thought that something like this is possible for German with its rather convoluted orthography. Thanks for doing this. It's without question useful to have such a template but maybe it should not be your top priority simply by the facts that, one, we have a VERY prolific German editor who only works on pronunciations (German has by far the best audio coverage on Wiktionary) and, two, because this is much more complex than pron templates in many, many other languages. I've now looked over the test cases and fixed some that immediately sprung out as wrong to me. Some further thoughts:
 * I think the module is supposed to generate broad transcriptions but it also generates ʔ (coincidentally, ʔ is the cause for quite a few of your failed testcases) which is traditionally not considered to be phonemic (there is however a group of modern minimal pair: Lehrerinnen <> Lehrer*innen (Gendersternchen); I personally don't know if this now suddenly makes ʔ phonemic...).
 * On the topic of broadness, I don't really know what the correct handling of [ɐ] <> [əʁ], [n̩] <> [ən] and [l̩] <> [əl] is but I'm inclined to think that they are allophones ([ɐ] not always being a free one) and that this is currently handled very inconsistently on Wiktionary (my opinion is that what we're currently using German broad transcriptions for is actually a middle ground: we use more symbols than there are phonemes but at least you don't get unpronounceable transcriptions that you'd first have to apply the non-free variations to). I started a discussion about this once but it fizzled out without any consensus reached.
 * Whether /ŋ/ is a phoneme is debatable (personally, I'd say no).
 * [ç] and [x] are positional allophones and not phonemic.
 * So all my points basically are that we have to be careful what we want to be part of the broad transcription. I'm not sure I can help much more than that unfortunately; I have not studied German phonology and some of the edge cases are indeed really edgy. &mdash; Fytcha〈 T | L | C 〉 19:13, 30 October 2022 (UTC)