Talk:tigre de papier

@Benwing2: Hi. I've noticed for some time now that the brackets added around /ə/ by {{fr-IPA}} are quite often wrong; for example, I'd say it's compulsory in this case. Removing the schwa here would be almost unpronounceable, imo. Per utramque cavernam 00:03, 4 November 2018 (UTC)

@Per utramque cavernam How would it be unpronounceable? As long as you pronounce the first schwa, omitting the second one is easy. If you omit the first schwa, then yes the second one is mandatory. As for handling this automatically, I have no idea how to do that; the algorithm I implemented is clearly wrong in many cases but I'm not sure if there's a better one. If you want to manually insert schwas, use ĕ to indicate a mandatory schwa. Benwing2 (talk) 05:44, 4 November 2018 (UTC)

@Benwing2: Sorry, I don't know what I was smoking last night. Indeed, if the first schwa is left removing the second is all right, at least theoretically.

It still sounds a bit unnatural to my ears though. If anything, I'd go the other way, and write/say /ti.ɡʁ(ə) də pa.pje/.

Thanks for the tip about adding a mandatory schwa; the result is now /ti.ɡʁə də pa.pje/, which is fine. However, could make your algorithm go a step further? When the template knows where the mandatory schwa is, could it recalculate the optionality of the other one, and arrive at what I was writing above, i.e. /ti.ɡʁ(ə) də pa.pje/? Per utramque cavernam 10:17, 4 November 2018 (UTC)

@Per utramque cavernam Sure, I will see about doing that. Benwing2 (talk) 18:07, 4 November 2018 (UTC)

I looked into the code and the relevant section is this:

	--6. make final schwa optional after two consonants except obstruent + approximant
	--   and [lmn] + ʁ
	text = rsub(text, '(' .. cons_c .. ')(' .. '%.?' .. ')(' .. cons_c .. ')ə⁀',
		function(a, dot, b)
			return a .. dot .. b .. (
				rfind(a, '[bdfɡkpstvzʃʒ]') and rfind(b, '[mnlʁwj]') and 'ə'
				or rfind(a, '[lmn]') and b == 'ʁ' and 'ə' or '(ə)') .. '⁀'
		end)

This handles final schwa after two consonants, which is made optional except in cases like [gʁ], [tm], etc. Note that this only applies when the next word begins with a consonant; preceding rules handle the case of final schwa in absolutely-final position and when the next word begins with a vowel. Should this rule be changed more generally? I'm aware that in fast speech, [l] and [ʁ] often get elided entirely in cases such as 'table verte' [tab vɛʁt] and 'tigre jaune' [tig ʒon], but I'm not sure whether we should indicate that. (What happens in fast speech with e.g. 'rythme vite'?) Benwing2 (talk) 18:27, 4 November 2018 (UTC)