This is a private module sandbox of Abuccts, for their own experimentation. Items in this module may be added and removed at Abuccts's discretion; do not rely on this module's stability.
--[[
This module implements the template {{ru-IPA}}.
Author: Originally Wyang; rewritten by Benwing; additional contributions
from Atitarev and a bit from others
FIXME:
1. (DONE) Geminated /j/ from -йя-: treat as any other gemination, meaning it
may not always be pronounced geminated. Currently we geminate it very late,
after all the code that reduces geminates. Should be done earlier and
places that include regexps with /j/ should be modified to also include
the gemination marker ː. Words with йя: аллилу́йя, ауйяма, ва́йя, ма́йя,
папа́йя, парано́йя, пира́йя, ра́йя, секво́йя, Гава́йям.
2. (DONE) Should have geminated jj in йе (occurs in e.g. фойе́). Should work
with gem=y (see FIXME #1). Words with йе: фойе́, колба Эрленмейера, скала
Айерс, Айерс-Рок, йети, Кайенна, конве́йер, конвейерный, сайентология,
фейерверк, Гава́йев. Note also Гава́йи with йи.
3. (DONE, CINEMANTIQUE OK WITH FIXES) In Асунсьо́н and Вьентья́н, put a syllable
break after the н and before consonant + /j/. Use the perm_syl_onset
mechanism or at least the code that accesses that mechanism. Should
possibly do this also in VCʲj and V‿Cʲj and VCj and V‿Cj sequences;
ask Cinemantique if this makes sense.
4. (DONE, CINEMANTIQUE OK WITH FIXES) Fix non-palatal е in льстец. Other
words that will be affected (and probably wrong): льви́ца, львя́тник,
льняно́й, льстить, льди́на, львиный, manual pronunciation given as lʲvʲit͡sə
and lʲvʲɵnək. Ask Cinemantique.
5. (DONE, CINEMANTIQUE SAYS NO IT DOESN'T) In львёнок, rendered as ˈlʲvɵnək
instead of ˈlʲvʲɵnək. Apparently same issue as льстец, having to do with
ь in beginning. This apparently has to do with the "assimilative
palatalization of consonants when followed by front vowels" code, which
blocks palatalization when the syllable begins with a cluster with a hard
sign, or a soft sign followed by a consonant. Then "retraction of front
vowels in syllables blocking assimilative palatalization" converts e to ɛ
and i to y in such cases of blocked palatalization (not needed for žcš,
which are handled by phon_respellings). Ask Cinemantique if this whole
business makes any sense.
6. (DONE) In prefixes/suffixes like -ин, treat single syllable word as
unstressed. Also support tilde to force unstressed syllable.
7. (DONE) In ни́ндзя, дз becomes palatal and н should palatal-assimilate to it.
8. (DONE) In под сту́лом, should render as pɐt͡s‿ˈstuləm when actually renders as
pɐˈt͡s‿stuləm. Also occurs in без ша́пки (bʲɪˈʂ‿ʂapkʲɪ instead of
bʲɪʂ‿ˈʂapkʲɪ); has something to do with ‿. Similarly occurs in
не к ме́сту, which should render as nʲɪ‿k‿ˈmʲestʊ, and от я́блони, which
should render as ɐt‿ˈjæblənʲɪ.
9. (STILL UNCLEAR) In собра́ние, Anatoli renders it as sɐˈbranʲɪ(j)ə with
optional (j). Ask him when this exactly applies. Does it apply in all ɪjə
sequences? Only word-finally? Also ijə?
10. (DONE) убе́жищa renders as ʊˈbʲeʐɨɕːʲə instead of ʊˈbʲeʐɨɕːə; уда́ча
similarly becomes ʊˈdat͡ɕʲə instead of ʊˈdat͡ɕə.
10a. (DONE) Remove the "offending clause" just mentioned, labeled FIXME (10a),
and fix it as the comment above it describes.
10b. (DONE) Remove the clause labeled "FIXME (10b)".
10c. (DONE) Investigate the clause labeled "FIXME (10c)". This relates to
FIXME #9 above concerning собра́ние.
10d. (DONE, NEEDS TESTING) Investigate the clause labeled "FIXME (10d)"
and apply the instructions there about removing a line and seeing
whether anything changes.
11. (DONE) тро́лль renders with geminated final l, and with ʲ on wrong side of
gemination (ːʲ instead of ʲː); note how this also occurs above in -ɕːʲə
from убе́жищa. (This issue with тро́лль will be masked by the change to
generally degeminate l; use фуррь; note also галльский.)
12. (DONE, NEEDS TESTING) May be additional errors with gemination in
combination with explicit / syllable boundary, because of the code
expecting that syllable breaks occur in certain places; should probably
rewrite the whole gemination code to be less fragile and not depend on
exactly where syllable breaks occur in consonant clusters, which it does
now (might want to rewrite to avoid the whole business of breaking by
syllable and processing syllable-by-syllable).
13. Many assimilations won't work properly with an explicit / syllable
boundary.
14. (DONE, ASK WYANG FOR ITS PURPOSE) Eliminate pal=y. Consider first asking
Wyang why this was put in originally.
15. (DONE) Add test cases: Цю́рих, от а́ба, others.
15a. (DONE) Add test cases: фуррь, по абази́ну (for assimilation of schwas
across ‿)
15b. (DONE) Add test case англо-норма́ннский (to make sure degemination of нн
occurs when not between vowels), multi-syllable word ending in a geminate:
ато́лл (not so good because лл always degeminated), коло́сс, Иоа́нн (good
because of нн), ха́ос, эвфеми́зм, хору́гвь (NOTE: ruwikt claims гв is voiced,
I doubt it, ask Cinemantique), наря́д на ку́хню (non-devoicing of д before
н in next word, ask Cinemantique about this, does it also apply to мрл?),
ко̀е-кто́
16. (DONE, ADDED SPECIAL HACK; REMOVED WITH NEW FINAL-Е CODE, SHOULD HANDLE
THROUGH pos=pro; DOESN'T HAVE ANYTHING TO DO WITH SECONDARY STRESS ON О)
Caused a change in ко̀е-кто́, perhaps because I rewrote code that accepted
an acute or circumflex accent to also take a grave accent. See how кое is
actually pronounced here and take action if needed. (ruwiki claims кое is
indeed pronounced like кои, ask Cinemantique what the rule for final -е
is and why different in кое vs. мороженое, anything to do with secondary
stress on о?)
17. (DONE) Rewrote voicing/devoicing assimilation; should make assimilation of
эвфеми́зм automatic and not require phon=.
18. (DONE) Removed redundant fronting-of-a code near end; make sure this
doesn't change anything.
19. (DONE, ANSWER IS YES) do сь and зь assimilate before шж, and
if so do they become ɕʑ? Ask Cinemantique.
20. (DONE) Add pos= to handle final -е. Possibilities appear to be neut
(neuter noun), adj (adjective, autodetected whether singular or plural),
comp (comparative), pre (prepositional case), adv (adverb), verb or v (2nd
plural verb forms).
21. (DONE, DEVOICE UNLESS NEXT WORD BEGINS WITH VOICED OBSTRUENT OR V+VOICED
OBSTRUENT) Figure out what to do with devoicing or non-devoicing before
mnrlv vowel. Apparently non-devoicing before vowel is only in fast speech
with a close juncture and Anatoli doesn't want that; but what about before
the consonants?
22. (DONE) Figure out what to do with fronting of a and u after or between
soft consonants, esp. when triggered by a following soft consonant with
optional or compulsory assimilation. Probably the correct thing to do
in the case of optional assimilation is to give two pronunciations
separated by commas, one with non-front vowel + hard consonant, the
other with front vowel + soft consonant.
23. (DONE, OK) Implement compulsory assimilation of xkʲ; ask Cinemantique to
make sure this is correct.
24. (DONE, BUT ANATOLI THINKS CONJUNCTION A MIGHT NOT BE REDUCED) Add а to
list of unstressed particles, but only recognize it and о (and perhaps all
the others) when not followed by a hyphen; then fix unnecessary cases with
о̂ (look at tracking cflex category) and the various hacks used in а ведь,
а сейчас, а то, а не то, а также, а как же; will need to add а̂ to а капелла
and possibly elsewhere; use different-pron tracking to catch this.
25. (DONE) Add / before цз, чж in Chinese words to ensure syllable boundary in
right place; ensure that this doesn't mess things up when occurring at
beginning of word or whatever.
26. (DONE) Rule on voicing assimilation before v: It says in Chew "A
Computational Phonology of Russian" that v is an obstruent before
obstruents and a sonorant before sonorants, i.e. v triggers voicing
assimilation if followed by an obstruent; verify that our code works this
way.
27. (DONE, NEEDS TESTING) Implement _ to block all assimilation; probably this
will happen automatically and we just need to remove the _ at the end.
28. (NOT DONE, NOT CORRECT) Change unstressed palatal o to have values like
regular o, for words like майора́т, Ога́йо, Йоха́ннесбург
29. (DONE) If we need partial reduction of non-final е/я to instead of ,
one way is to use another diacritic, e.g. dot-under; or use a spelling
like ьа.
30. (DONE) BUG: воѐнно-морско́й becomes without , must be
due to ѐ being a composed character (may be a bug in the translit module;
add a test case).
31. In в Япо́нии, в Евро́пе, the initial should be required not optional.
32. (DONE) Should be possible to write п(ь)я́нка, скам(ь)я́ and get optional
palatalization.
33. (CODE PRESENT BUT NOT COMPLETED) Final unstressed -е that becomes
should become when not followed by end of utterance or pause.
34. In То́гане (phon=То́ганэ), final -э should be pronounced . Should apply
in general to -э after paired consonants, but not to e.g. се́рдце.
35. (DONE) тц,дц,тч,дч shoud be always-geminated by default.
36. (DONE) treat ! and ? as separate words so we don't have issues with
word-final -е before them.
37. (DONE) Distinguish stress accents from other accents.
38. т(ь)ся not directly after the stress should be optionally geminated.
39. (DONE) нра̀вственно-эти́ческий should have optional not mandatory gemination
of нн.
40. (DONE) Make дц in -дцат- be optionally-geminated, for words like
одиннадцать, двадцать, тридцатый, etc.
41. (DONE) Don't show grave accents in annotations (but do in phon=).
42. (DONE) -чш- (as in лучший) should be pronounced as -тш-.
43. (DONE) Fix fronting of in two syllables in a row.
44. (DONE) Add pos=imp for imperatives, use it to treat -ться differently from
infinitives.
45. (DONE) CFLEX should not be treated as stress for the purpose of determining
whether written а reduces to or .
46. (DONE) Fix ьт, ьс sequences (esp. in imperatives) and make
palatalization of labials optional in ь (again esp. in
imperatives).
47. (DONE) Optional palatalization of -ся should apply only to -лся, not always.
48. (DONE) Reduction of стл -> сл should apply only in стлив, not always.
49. (DONE) Convert счит -> щит by default, as with счёт.
50. (DONE) Don't require that m_ru_translit.apply_tr_fixes() be called prior
to ipa(), but include an argument so that text transformed this way can
be passed in.
51. (DONE) pos=X/Y and gem=X/Y should require same number of elements as actual
words rather than counting phonetically-joined words.
]]
local ut = require("Module:utils")
local com = require("Module:ru-common")
local m_ru_translit = require("Module:ru-translit")
local strutils = require("Module:string utilities")
local export = {}
local u = mw.ustring.char
local rfind = mw.ustring.find
local rsubn = mw.ustring.gsub
local rmatch = mw.ustring.match
local rsplit = mw.text.split
local ulower = mw.ustring.lower
local uupper = mw.ustring.upper
local usub = mw.ustring.sub
local ulen = mw.ustring.len
local remove_grave_accents_from_phonetic_respelling = true -- Anatoli's desired value
-- version of rsubn() that discards all but the first return value
local function rsub(term, foo, bar)
local retval = rsubn(term, foo, bar)
return retval
end
-- apply rsub() repeatedly until no change
local function rsub_repeatedly(term, foo, bar)
while true do
local new_term = rsub(term, foo, bar)
if new_term == term then
return term
end
term = new_term
end
end
-- Clone parent's args while also assigning nil to empty strings.
local function clone_args(frame)
local args = {}
for pname, param in pairs(frame:getParent().args) do
if param == "" then args = nil
else args = param
end
end
return args
end
-- If enabled, compare this module with new version of module in
-- Module:User:Benwing2/ru-pron to make sure all pronunciations are the same.
-- To check for differences, go to Wiktionary:Tracking/ru-pron/different-pron
-- and look at what links to the page.
local test_new_ru_pron_module = false
-- If enabled, do new code for final -е; else, the old way
local new_final_e_code = true
-- If enabled, do special case for final -е not before a pause
local final_e_non_pausal = false
local AC = u(0x0301) -- acute = ́
local GR = u(0x0300) -- grave = ̀
local CFLEX = u(0x0302) -- circumflex = ̂
local DUBGR = u(0x030F) -- double grave = ̏
local DOTABOVE = u(0x0307) -- dot above = ̇
local DOTBELOW = u(0x0323) -- dot below = ̣
local TEMPCFLEX = u(0xFFF1) -- placeholder to be converted to a circumflex
local vow = 'aeiouyɛəäạëöü'
local ipa_vow = vow .. 'ɐɪʊɨæɵʉ'
local vowels, vowels_c = '', '()'
-- No need to include DUBGR here because we rewrite it to CFLEX very early
local acc = AC .. GR .. CFLEX .. DOTABOVE .. DOTBELOW
local accents = ''
local stress_accents = ''
local perm_syl_onset = ut.list_to_set({
'spr', 'str', 'skr', 'spl', 'skl',
-- FIXME, do we want sc?
'sp', 'st', 'sk', 'sf', 'sx', 'sc',
'pr', 'br', 'tr', 'dr', 'kr', 'gr', 'fr', 'vr', 'xr',
'pl', 'bl', 'kl', 'gl', 'fl', 'vl', 'xl',
-- FIXME, do we want the following? If so, do we want vn?
'ml', 'mn',
-- FIXME, dž is now converted to ĝž, which will have a syllable
-- boundary in between
'šč', 'dž',
})
-- FIXME: Consider changing ӂ internally to ʑ to match ɕ (it is used externally
-- in e.g. дроӂӂи (pronunciation spelling of дрожжи)
local translit_conv = {
= 't͡s', = 't͡ɕ', = 't͡ʂ',
= 'ɡ', = 'd͡ʐ',
= 'd͡z', = 'd͡ʑ', = 'ʑ',
= 'ʂ', = 'ʐ'
}
local translit_conv_j = {
= 't͡sʲ',
= 'd͡zʲ'
}
local allophones = {
= { 'a', 'ɐ', 'ə' },
= { 'e', 'ɪ', 'ɪ' },
= { 'i', 'ɪ', 'ɪ' },
= { 'o', 'ɐ', 'ə' },
= { 'u', 'ʊ', 'ʊ' },
= { 'ɨ', 'ɨ', 'ɨ' },
= { 'ɛ', 'ɨ', 'ɨ' },
= { 'a', 'ɪ', 'ɪ' },
= { 'a', 'ɐ', 'ə' },
= { 'e', 'ɪ', 'ɪ' },
= { 'ɵ', 'ɪ', 'ɪ' },
= { 'u', 'ʊ', 'ʊ' },
= { 'ə', 'ə', 'ə' },
}
local devoicing = {
= 'p', = 't', = 'k',
= 's', = 'f',
= 'š', = 'x',
= 'c', = 'č', = 'ĉ',
= 'ɕ',
}
local voicing = {
= 'b', = 'd', = 'g',
= 'z', = 'v',
= 'ž', = 'ĵ', = 'ǰ', = 'ĝ',
= 'ɣ', = 'ӂ'
}
local iotating = {
= 'ä',
= 'ë',
= 'ö',
= 'ü'
}
local retracting = {
= 'ɛ',
= 'y',
}
local fronting = {
= 'æ',
= 'ʉ',
= 'ʉ',
}
-- Prefixes that we recognize specially when they end in a geminated
-- consonant. The first element is the result after applying voicing/devoicing,
-- gemination and other changes. The second element is the original spelling,
-- so that we don't overmatch and get cases like Поттер. We check for these
-- prefixes at the beginning of words and also preceded by ne-, po- and nepo-.
local geminate_pref = {
--'abː', --'adː',
{'beː', 'be'},
--'braomː',
{'ː', 'v'},
{'voː', 'vo'},
{'iː', 'i'},
--'^inː',
{'kontrː', 'kontr'},
{'superː', 'super'},
{'tranː', 'trans'},
{'naː', 'nad'},
{'niː', 'ni'},
{'oː', 'ot'}, --'^omː',
{'oː', 'ob'},
{'obeː', 'obe'},
{'poː', 'pod'},
{'preː', 'pred'}, --'^paszː', '^pozː',
{'raː', 'ra'},
{'ː', ''}, -- ž on right for жжёт etc., ш on left for США
{'meː', 'mež'},
{'če?reː', 'če?re'},
-- certain double prefixes involving ra-
{'predraː', 'predra'},
{'bezraː', 'bezra'},
{'naraː', 'nara'},
{'vraː', 'vra'},
{'doraː', 'dora'},
-- '^sverxː', '^subː', '^tröxː', '^četyröxː',
}
local sztab = { s='cs', z='ĵz' }
local function ot_pod_sz(pre, sz)
return pre .. sztab
end
local phonetic_subs = {
{'h', 'ɣ'},
{'šč', 'ɕː'}, -- conversion of šč to geminate
-- the following group is ordered before changes that affect ts
{'nsk', 'n(t)sk'},
{'ssk', 'sck'},
-- Add / before цз, чж sequences (Chinese words) and assimilate чж
{'cz', '/cz'},
{'čž', '/ĝž'},
-- main changes for affricate assimilation of + sibilant, including ts;
-- we either convert to "short" variants t͡s, d͡z, etc. or to "long" variants
-- t͡ss, d͡zz, etc.
-- 1. т с, д з across word boundary, also т/с, д/з with explicitly written
-- slash, use long variants.
{'(ʹ?+)s', 'c%1s'},
{'(ʹ?+)z', 'ĵ%1z'},
-- 2. тс, дз + vowel use long variants.
{'(ʹ?)s(j?' .. vowels .. ')', 'c%1s%2'},
{'(ʹ?)z(j?' .. vowels .. ')', 'ĵ%1z%2'},
-- 3. тьс, дьз use long variants.
{'ʹs', 'cʹs'},
{'ʹz', 'ĵʹz'},
-- 4. word-initial от-, под- use long variants because there is
-- a morpheme boundary.
{'(⁀o' .. accents .. '?)t()', ot_pod_sz},
{'(⁀po' .. accents .. '?)d()', ot_pod_sz},
-- 5. other тс, дз use short variants.
{'s', 'c'},
{'z', 'ĵ'},
-- 6. тш, дж always use long variants (FIXME, may change)
{'(ʹ?*)š', 'ĉ%1š'},
{'(ʹ?*)ž', 'ĝ%1ž'},
-- 7. soften palatalized hard hushing affricates resulting from the previous
{'ĉʹ', 'č'},
{'ĝʹ', 'ǰ'},
-- changes that generate ɕː and ɕč through assimilation:
-- зч and жч become ɕː, as does сч at the beginning of a word and in the
-- sequence счёт when not following (подсчёт); else сч becomes ɕč
-- (отсчи́тываться), as щч always does (рассчитáть written ращчита́ть)
{'sč', 'čɕː'},
{'ɕːč', 'ɕč'},
{'č', 'ɕː'},
{'ɕː?', 'ɕː'},
{'⁀sč', '⁀ɕː'},
{'sč(j?' .. accents .. '?)t', 'ɕː%1t'},
{'sč', 'ɕč'},
-- misc. changes for assimilation of + sibilants and affricates
{'c', 'sc'},
{'()()', '%1%2'},
-- -дцат- (in numerals) has optionally-geminated дц
{'dca(' .. accents .. '?)t', 'c(c)a%1t'},
-- дц, тц, дч, тч + vowel always remain geminated, so mark this with ˑ;
-- if not followed by a vowel, as in e.g. путч, use normal gemination
-- (it will normally be degeminated)
{'()(' .. vowels .. ')', '%1ˑ%2'},
{'()', '%1%1'},
-- the following is ordered before the next one, which applies assimilation
-- of to щ (including across word boundaries)
{'nɕ', 'nɕ'},
-- and ь before soft affricates , including across word
-- boundaries; note that the common sequence сч has already been handled
{'ʹ?(*)', 'ɕ%1'},
-- reduction of too many ɕ's, which can happen from the previous
{'ɕɕː', 'ɕː'},
-- assimilation before and ь before щ
{'ʹ?(*)ɕ', 'č%1ɕ'},
-- assimilation of and ь before
{'(*)š', 'š%1š'},
{'(*)ž', 'ž%1ž'},
{'ʹ(*)š', 'ɕ%1š'},
{'ʹ(*)ž', 'ӂ%1ž'},
-- assimilation of ь before с (in imperatives esp. before ся)
{'ʹs()', 'sˑ%1'},
-- assimilation of ь before т (e.g. in imperatives esp. before те)
{'ʹt()', 'tˑ%1'},
-- optional palatalization of palatalized labials before another consonant
-- in (esp. in imperatives before -те, -ся)
-- FIXME, perhaps we should either generalize this or restrict it only
-- to imperatives
{'()ʹ()', '%1(ʹ)%2'},
{'sverxi', 'sverxy'},
{'stʹd', 'zd'},
-- this will often become degeminated
{'tʹd', 'dd'},
-- loss of consonants in certain clusters
{'()g', '%1g'},
{'zdn', 'zn'},
{'lnc', 'nc'},
{'t(li' .. accents .. '?v)', 's%1'},
{'tn', 'sn'},
-- backing of /i/ after hard consonants in close juncture
{'()⁀‿⁀i', '%1⁀‿⁀y'},
}
local cons_assim_palatal = {
-- assimilation of tn, dn, sn, zn, st, zd, nč, nɕ is handled specially
compulsory = ut.list_to_set({'ntʲ', 'ndʲ', 'xkʲ',
'csʲ', 'ĵzʲ', 'ncʲ', 'nĵʲ'}),
optional = ut.list_to_set({'slʲ', 'zlʲ', 'nsʲ', 'nzʲ',
'mpʲ', 'mbʲ', 'mfʲ', 'fmʲ'})
}
-- words which will be treated as accentless (i.e. their vowels will be
-- reduced), and which will liaise with a preceding or following word;
-- this will not happen if the words have an accent mark, cf.
-- по́ небу vs. по не́бу, etc.
local accentless = {
-- class 'pre': particles that join with a following word
pre = ut.list_to_set({'bez', 'bliz', 'v', 'vo', 'da', 'do',
'za', 'iz', 'iz-pod', 'iz-za', 'izo', 'k', 'ko', 'mež',
'na', 'nad', 'nado', 'ne', 'ni', 'ob', 'obo', 'ot', 'oto',
'pered', 'peredo', 'po', 'pod', 'podo', 'pred', 'predo', 'pri', 'pro',
's', 'so', 'u', 'čerez'}),
-- class 'prespace': particles that join with a following word, but only
-- if a space (not a hyphen) separates them; hyphens are used here
-- to spell out letters, e.g. а-эн-бэ́ for АНБ (NSA = National Security
-- Agency) or о-а-э́ for ОАЭ (UAE = United Arab Emirates)
prespace = ut.list_to_set({'a', 'o'}),
-- class 'post': particles that join with a preceding word
post = ut.list_to_set({'by', 'b', 'ž', 'že', 'li', 'libo', 'lʹ', 'ka',
'nibudʹ', 'tka'}),
-- class 'posthyphen': particles that join with a preceding word, but only
-- if a hyphen (not a space) separates them
posthyphen = ut.list_to_set({'to'}),
}
-- Pronunciation of final unstressed -е, depending on the part of speech and
-- exact ending. Also used for pronunciation of -ться in imperatives vs.
-- infinitives.
--
-- Endings:
-- oe = -ое
-- ve = any other vowel plus -е (FIXME, may have to split out -ее)
-- je = -ье
-- softpaired = soft paired consonant + -е
-- hardsib = hard sibilant (ц, ш, ж) + -е
-- softsib = soft sibilant (ч, щ) + -е
--
-- Parts of speech:
-- def = default used in absence of pos
-- n/noun = neuter noun in the nominative/accusative singular (but not ending
-- in adjectival -ое or -ее; those should be considered as adjectives)
-- pre = prepositional case singular
-- dat = dative case singular (treated same as prepositional case singular)
-- voc = vocative case (currently treated as 'mid')
-- nnp = noun nominative plural in -е (гра́ждане, боя́ре, армя́не); not
-- adjectival plurals in -ие or -ые, including adjectival nouns
-- (да́нные, а́вторские)
-- inv = invariable noun or other word (currently treated as 'mid')
-- a/adj = adjective or adjectival noun (typically either neuter in -ое or
-- -ее, or plural in -ие, -ые, or -ье, or short neuter in unpaired
-- sibilant + -е)
-- c/com = comparative (typically either in -ее or sibilant + -е)
-- adv = adverb
-- p = preposition (treated same as adverb)
-- v/vb/verb = finite verbal form (usually 2nd-plural in -те), but not
-- imperatives (use pos=imp) and not participle forms, which should be
-- treated as adjectives
-- pro = pronoun (кое-, какие-, ваше, сколькие)
-- num = number (двое, трое, обе, четыре; currently treated as 'mid')
-- pref = prefix (treated as 'high' because integral part of word)
-- hi/high = force high values ( or )
-- mid = force mid values ( or )
-- lo/low/schwa = force low, really schwa, values ()
--
-- Possible values:
-- 1. ə , e , i after a vowel or soft consonant
-- 2. ə or y after a hard sibilant
--
-- If a part of speech doesn't have an entry for a given type of ending,
-- it receives the default value. If a part of speech's entry is a string,
-- it's an alias for another way of specifying the same part of speech
-- (e.g. n=noun).
local pos_properties = {
def={oe='ə', ve='e', je='e', softpaired='e', hardsib='y', softsib='e', tsjapal='n'},
noun={oe='ə', ve='e', je='e', softpaired='e', hardsib='ə', softsib='e'},
n='noun',
pre={oe='e', ve='e', softpaired='e', hardsib='y', softsib='e'},
dat='pre',
voc='mid',
nnp={softpaired='e'}, -- FIXME, not sure about this
inv='mid', --FIXME, not sure about this (e.g. вице-, кофе)
adj={oe='ə', ve='e', je='ə'}, -- FIXME: Not sure about -ее, e.g. neut adj си́нее; FIXME, not sure about short neuter adj, e.g. похо́же from похо́жий, дорогосто́яще from дорогосто́ящий, should this be treated as neuter noun?
a='adj',
com={ve='e', hardsib='y', softsib='e'},
c='com',
adv={softpaired='e', hardsib='y', softsib='e'},
p='adv', --FIXME, not sure about prepositions
verb={softpaired='e'},
v='verb',
vb='verb',
-- Imperatives like other verbs except that final -ться is palatalized
imp={softpaired='e', tsjapal='y'},
impv='imp',
pro={oe='i', ve='i'}, --FIXME, not sure about ваше, сколькие, какие-, кое-
num='mid', --FIXME, not sure about обе
pref='high',
-- forced values
high={oe='i', ve='i', je='i', softpaired='i', hardsib='y', softsib='i'},
hi='high',
mid={oe='e', ve='e', je='e', softpaired='e', hardsib='y', softsib='e'},
low={oe='ə', ve='ə', je='ə', softpaired='ə', hardsib='ə', softsib='ə'},
lo='low',
schwa='low'
}
local function ine(x)
return x ~= "" and x or nil
end
local function track(page)
local m_debug = require("Module:debug")
m_debug.track("ru-pron/" .. page)
return true
end
-- remove accents that we don't want to appear in the phonetic respelling
function phon_respelling(text, remove_grave)
text = rsub(text, '', '')
-- Remove grave accents from annotations but maybe not from phonetic respelling
if remove_grave then
text = com.remove_grave_accents(text)
end
return text
end
function export.ru_IPA(frame)
local args = clone_args(frame)
local text = args or args.phon or mw.title.getCurrentTitle().text
local origtext, transformed_text = m_ru_translit.apply_tr_fixes(text,
args.noadj, args.noshto)
local maintext = export.ipa(transformed_text, args.adj, args.gem, args.bracket or "y", args.pos, "transformed")
if args.raw then
return maintext
else
local anntext = (args.ann == "y" and "'''" .. phon_respelling(text, "remove grave") .. "''': " or
args.ann and "'''" .. args.ann .. "''': " or
"")
local lang = require("Module:languages").getByCode("ru")
maintext = require("Module:IPA").format_IPA_full { lang = lang, items = {{pron = maintext}} }
local respelling
if args.phon then
respelling = args.phon
elseif origtext ~= transformed_text then
respelling = transformed_text
end
local respelling_text = ""
if respelling then
respelling = phon_respelling(respelling, remove_grave_accents_from_phonetic_respelling)
respelling_text = respelling and " (''phonetic respelling'': " .. require("Module:script utilities").tag_text(respelling, lang) .. ")"
end
return anntext .. maintext .. respelling_text
end
end
-- Return the actual IPA corresponding to Cyrillic text. ADJ, GEN, BRACKET
-- and POS are as in ]. If IS_TRANFORMED is true, the text
-- has already been passed through m_ru_translit.apply_tr_fixes(); otherwise,
-- this will be done.
function export.ipa(text, adj, gem, bracket, pos, is_transformed)
local new_module_result
local debug = {}
if type(text) == 'table' then
text, adj, gem, bracket, pos = (ine(text.args.phon) or ine(text.args)), ine(text.args.adj), ine(text.args.gem), ine(text.args.bracket), ine(text.args.pos)
if not text then
text = mw.title.getCurrentTitle().text
end
end
table.insert(debug, text)
local origtext, transformed_text = m_ru_translit.apply_tr_fixes(text)
text = transformed_text
table.insert(debug, text)
gem = gem or ""
-- If a multipart gemination spec, split into components.
if rfind(gem, "/") then
gem = rsplit(gem, "/")
for i=1,#gem do
gem = usub(gem, 1, 1)
end
else
gem = usub(gem, 1, 1)
end
-- Verify that gem (or each part of multipart gem) is recognized
for _, g in ipairs(type(gem) == "table" and gem or {gem}) do
if g ~= "" and g ~= "y" and g ~= "o" and g ~= "n" then
error("Unrecognized gemination spec '" .. g .. ": Should be y, yes, o, opt, n, no, or empty")
end
end
bracket = ine(usub(bracket or '', 1, 1))
if bracket == 'n' then
bracket = nil
end
pos = pos or "def"
-- If a multipart part of speech, split into components, and convert
-- each blank component to the default.
if rfind(pos, "/") then
pos = rsplit(pos, "/")
for i=1,#pos do
if pos == "" then
pos = "def"
end
end
end
-- Verify that pos (or each part of multipart pos) is recognized
for _, p in ipairs(type(pos) == "table" and pos or {pos}) do
if not pos_properties then
error("Unrecognized part of speech '" .. p .. "': Should be n/noun/neut, a/adj, c/com, pre, dat, adv, inv, voc, v/verb, pro, hi/high, mid, lo/low/schwa or omitted")
end
end
text = ulower(text)
local combined_gem = type(gem) == "table" and table.concat(gem, "/") or gem
if combined_gem ~= "" then
track("gem")
track("gem/" .. combined_gem)
end
if adj then
track("adj")
end
-- don't include h here because we allow it as a legitimate alternative
-- for ɣ. Include vowels with all of the accents that have special meaning
-- for this module. (FIXME, maybe should also include double-grave accents,
-- although probably not used anywhere.)
if rfind(text, "") then
track("latin-text")
end
if rfind(text, "ч") then
track("sch")
end
if rfind(text, "ч") then
track("shch")
end
if rfind(text, CFLEX) then
track("cflex")
end
if rfind(text, DUBGR) then
track("dubgr")
end
text = rsub(text, "``", DUBGR)
text = rsub(text, "`", GR)
text = rsub(text, "@", DOTABOVE)
text = rsub(text, "%^", CFLEX)
text = rsub(text, DUBGR, CFLEX)
-- translit doesn't always convert э to ɛ (depends on whether a consonant
-- precedes), so do it ourselves before translit
text = rsub(text, 'э', 'ɛ')
-- vowel + йе should have double jj, but the translit module will translit
-- it the same as vowel + е, so do it ourselves before translit
text = rsub(text, '(' .. com.opt_accent .. ')й()',
'%1йй%2')
-- transliterate and decompose Latin vowels with accents, recomposing
-- certain key combinations; don't include accent on monosyllabic ё, so
-- that we end up without an accent on such words. NOTE: Not clear we
-- need to be decomposing like this any more, although it is still
-- useful if the user supplies Latin text, which we allow (although
-- undocumented).
text = com.decompose(m_ru_translit.tr_after_fixes(text))
-- handle old ě (e.g. сѣдло́), and ě̈ from сѣ̈дла
text = rsub(text, 'ě̈', 'jo' .. AC)
text = rsub(text, 'ě', 'e')
-- handle sequences of accents (esp from ё with secondary/tertiary stress)
text = rsub(text, accents .. '+(' .. accents .. ')', '%1')
-- convert commas and en/en dashes to IPA foot boundaries
text = rsub(text, '%s*%s*', ' | ')
-- canonicalize multiple spaces
text = rsub(text, '%s+', ' ')
table.insert(debug, text)
-- Add primary stress to single-syllable words preceded or followed by
-- unstressed particle or preposition. Add "tertiary" stress to remaining
-- single-syllable words that aren't a particle, preposition, prefix or
-- suffix and don't already bear an accent (including force-reduction
-- accents, i.e. dot-above/dot-below); "tertiary stress" means a vowel is
-- treated as stressed for the purposes of vowel reduction but isn't
-- marked with a primary or secondary stress marker; we repurpose a
-- circumflex for this purpose. We need to preserve the distinction
-- between spaces and hyphens because (1) we only recognize certain
-- post-accentless particles following a hyphen (to distinguish e.g.
-- 'то' from '-то'); (2) we only recognize certain pre-accentless
-- particles preceding a space (to distinguish particles 'о' and 'а' from
-- spelled letters о and а, which should not be reduced); and (3) we
-- recognize hyphens for the purpose of marking unstressed prefixes and
-- suffixes.
local word = strutils.capturing_split(text, "(+)")
table.insert(debug, table.concat(word, "@"))
for i = 1, #word do
-- check for single-syllable words that need a stress; they must meet
-- the following conditions:
-- 1. must not be an accentless word, which is any of the following:
-- 1a. in the "pre" class, or
if not (accentless] or
-- 1b. in the "prespace" class if followed by space and another word, or
i < #word - 1 and accentless] and word == " " or
-- 1c. in the "post" class if preceded by another word and
-- not followed by a hyphen (this is because words like
-- ка and же are also used for spelling initialisms), or
i > 2 and accentless] and word ~= "-" or
-- 1d. in the "posthyphen" class preceded by a hyphen and another word
-- (and not followed by a hyphen, see 1c);
i > 2 and accentless] and word == "-" and word ~= "-") and
-- 2. must be one syllable;
ulen(rsub(word, '', '')) == 1 and
-- 3. must not have any accents (including dot-above, forcing reduction);
not rfind(word, accents) and
-- 4. must not be a prefix or suffix, identified by a preceding or trailing hyphen, i.e. one of the following:
-- 4a. utterance-initial preceded by a hyphen, or
not (i == 3 and word == "-" and word == "" or
-- 4b. non-utterance-initial preceded by a hyphen, or
i >= 3 and word == " -" or
-- 4c. utterance-final followed by a hyphen, or
i == #word - 2 and word == "-" and word == "" or
-- 4d. non-utterance-final followed by a hyphen;
i <= #word - 2 and word == "- ") then
-- OK, we have a stressable single-syllable word; either add primary
-- or tertiary stress:
-- 1. add primary stress if preceded or followed by an accentless word,
if (i > 2 and accentless] or
i > 2 and word == " " and accentless] or
i < #word - 1 and accentless] and word ~= "-" or
i < #word - 1 and word == "-" and accentless] and word ~= "-") then
word = rsub(word, vowels_c, '%1' .. AC)
-- 2. else add tertiary stress
else
word = rsub(word, vowels_c, '%1' .. CFLEX)
end
end
end
table.insert(debug, table.concat(word, "@"))
-- count number of words and make sure we have correct number of
-- gemination and part-of-speech specs if a multipart spec is given
local num_real_words = 0
for i = 1, #word do
if (i % 2) == 1 and word ~= "" then
num_real_words = num_real_words + 1
end
end
if type(gem) == "table" and #gem ~= num_real_words then
error("Number of gemination specs (" .. #gem .. ") should match number of words (" .. num_real_words .. ")")
end
if type(pos) == "table" and #pos ~= num_real_words then
error("Number of parts of speech (" .. #pos .. ") should match number of words (" .. num_real_words .. ")")
end
table.insert(debug, table.concat(word, "@"))
-- make unaccented prepositions and particles liaise with the following or
-- preceding word; in the process, fix up number of elements in gem/pos
-- tables so there's a single element for the combined word
local real_word_index = 0
for i = 1, #word do
if (i % 2) == 1 and word ~= "" then
real_word_index = real_word_index + 1
end
if i < #word - 1 and (accentless] or accentless] and word == " ") then
word = '‿'
if type(gem) == "table" then
table.remove(gem, real_word_index)
end
if type(pos) == "table" then
table.remove(pos, real_word_index)
end
elseif i > 2 and (accentless] and word ~= "-" or
accentless] and word == "-" and word ~= "-") then
word = '‿'
-- for unaccented words that liaise with the preceding word,
-- remove the gemination spec corresponding to the unaccented word
-- because the gemination in question is almost certainly in the
-- preceding word, but remove the POS spec corresponding to the
-- preceding word because it's the final -е of the unaccented word
-- that the POS will refer to
if type(gem) == "table" then
table.remove(gem, real_word_index)
end
if type(pos) == "table" then
table.remove(pos, real_word_index - 1)
end
end
end
table.insert(debug, table.concat(word, "@"))
-- rejoin words, convert hyphens to spaces and eliminate stray spaces
-- resulting from this
text = table.concat(word, "")
text = rsub(text, '+', ' ')
text = rsub(text, '^ ', '')
text = rsub(text, ' $', '')
-- add a ⁀ at the beginning and end of every word and at close juncture
-- boundaries; we will remove this later but it makes it easier to do
-- word-beginning and word-end rsubs
text = rsub(text, ' ', '⁀ ⁀')
text = rsub(text, '()', '⁀%1⁀')
text = '⁀' .. text .. '⁀'
text = rsub(text, '‿', '⁀‿⁀')
-- save original word spelling before respellings, (de)voicing changes,
-- geminate changes, etc. for implementation of geminate_pref
local orig_word = rsplit(text, " ", true)
-- insert or remove /j/ before so that palatal versions of these
-- vowels are always preceded by /j/ and non-palatal versions never are
-- (do this before the change below adding tertiary stress to final
-- palatal о):
-- (1) Non-palatal after always-hard шж (e.g. in брошю́ра, жю́ри)
-- despite the spelling (FIXME, should this also affect ?)
text = rsub(text, '()j()', '%2%3')
-- (2) Palatal after always-soft щчӂ and voiced variant ǰ (NOTE:
-- this happens before the change šč -> ɕː in phonetic_subs)
text = rsub(text, '()()', '%1j%2')
-- (3) ьо is pronounced as ьйо, i.e. like (possibly unstressed) ьё, e.g.
-- in Асунсьо́н
text = rsub(text, 'ʹo', 'ʹjo')
-- add tertiary stress to some final -о (this needs to be done before
-- eliminating dot-above, after adding ⁀, after adding /j/ before palatal о):
-- (1) after vowels, e.g. То́кио
text = rsub(text, '(' .. vowels .. accents .. '?o)⁀', '%1' .. CFLEX .. '⁀')
-- (2) when palatal, e.g. ра́нчо, га́учо, ма́чо, Ога́йо
text = rsub(text, 'jo⁀', 'jo' .. CFLEX .. '⁀')
-- eliminate dot-above, which has served its purpose of preventing any
-- sort of stress (needs to be done after adding tertiary stress to
-- final -о)
text = rsub(text, DOTABOVE, '')
-- eliminate dot-below (needs to be done after changes above that insert
-- j before after always-soft щчӂ)
text = rsub(text, 'ja' .. DOTBELOW, 'jạ')
if rfind(text, DOTBELOW) then
error("Dot-below accent can only be placed on я or palatal а")
end
text = adj and rsub(text, '(.́?)go(' .. AC .. '?)⁀', '%1vo%2⁀') or text
text = adj and rsub(text, '(.́?)go(' .. AC .. '?)sja⁀', '%1vo%2sja⁀') or text
function fetch_pos_property(i, ending)
local thispos = type(pos) == "table" and pos or pos
local chart = pos_properties
while type(chart) == "string" do -- handle aliases
chart = pos_properties
end
assert(type(chart) == "table")
local sub = chart or pos_properties
assert(sub)
return sub
end
-- Pos-specific handling of final -ться: palatalized if pos=imp, else not
-- (infinitives). If we have multiple parts of speech, we need to be
-- trickier, splitting by word.
local function final_tsja_processing(pron, i)
local tsjapal = fetch_pos_property(i, 'tsjapal')
if tsjapal == 'n' then
-- FIXME!!! Should these also pay attention to grave accents?
pron = rsub(pron, '́tʹ?sja⁀', '́cca⁀')
pron = rsub(pron, '()tʹ?sja⁀', '%1ca⁀')
end
return pron
end
if type(pos) == "table" then
--split by word and process each word
word = rsplit(text, " ", true)
for i = 1, #word do
word = final_tsja_processing(word, i)
end
text = table.concat(word, " ")
else
text = final_tsja_processing(text, 1)
end
--phonetic substitutions of various sorts
for _, phonsub in ipairs(phonetic_subs) do
text = rsub(text, phonsub, phonsub)
end
--voicing, devoicing
--NOTE: v before an obstruent assimilates in voicing and triggers voicing
--assimilation of a preceding consonant; neither happens before a sonorant
--1. absolutely final devoicing
text = rsub(text, '()(ʹ?⁀)$', function(a, b)
return devoicing .. b end)
--2. word-final devoicing before another word
text = rsub(text, '()(ʹ?⁀ ⁀)', function(a, b)
return devoicing .. b end)
--3. voicing/devoicing assimilation; repeat to handle recursive assimilation
while true do
local new_text = rsub(text, '()(*)', function(a, b)
return devoicing .. b end)
new_text = rsub(new_text, '()(*v?*)', function(a, b)
return voicing .. b end)
if new_text == text then
break
end
text = new_text
end
--re-notate orthographic geminate consonants
text = rsub(text, '()' .. '%1', '%1ː')
text = rsub(text, '()' .. '%(%1%)', '%1(ː)')
--rewrite iotated vowels
text = rsub(text, '(j*)()', function(a, b)
return a .. iotating end)
-- eliminate j after consonant and before iotated vowel (including
-- semi-reduced ạ)
text = rsub(text, '(/?)j()', '%1%2')
--split by word and process each word
word = rsplit(text, " ", true)
table.insert(debug, text)
for i = 1, #word do
local pron = word
-- Check for gemination at prefix boundaries; if so, convert the
-- regular gemination symbol ː to a special symbol ˑ that indicates
-- we always preserve the gemination unless gem=n. We look for
-- certain sequences at the beginning of a word, but make sure that
-- the original spelling is appropriate as well (see comment above
-- for geminate_pref).
if rfind(pron, 'ː') then
local orig_pron = orig_word
local deac = rsub(pron, accents, '')
local orig_deac = rsub(orig_pron, accents, '')
for _, gempref in ipairs(geminate_pref) do
local newspell = gempref
local oldspell = gempref
-- FIXME! The rsub below will be incorrect if there is
-- gemination in a joined preposition or particle
if rfind(orig_deac, '⁀' .. oldspell) and rfind(deac, '⁀' .. newspell) or
rfind(orig_deac, '⁀ne' .. oldspell) and rfind(deac, '⁀ne' .. newspell) then
pron = rsub(pron, '(⁀*)ː', '%1ˑ')
end
end
end
--degemination, optional gemination
local thisgem = type(gem) == "table" and gem or gem
if thisgem == 'y' then
-- leave geminates alone, convert ˑ to regular gemination; ˑ is a
-- special gemination symbol used at prefix boundaries that we
-- remove only when gem=n, else we convert it to regular gemination
pron = rsub(pron, 'ˑ', 'ː')
elseif thisgem == 'o' then
-- make geminates optional, except for ɕӂ, also ignore left paren
-- in (ː) sequence
pron = rsub(pron, '()', '%1(ː)')
elseif thisgem == 'n' then
-- remove gemination, except for ɕӂ
pron = rsub(pron, '()', '%1')
else
-- degeminate l's
pron = rsub(pron, '(l)ː', '%1')
-- preserve gemination between vowels immediately after the stress,
-- special gemination symbol ˑ also remains, ɕӂ remain geminated,
-- žn remain geminated between vowels even not immediately after
-- the stress, n becomes optionally geminated when after but not
-- immediately after the stress, ssk and zsk remain geminated
-- immediately after the stress, else degeminate; we signal that
-- gemination should remain by converting to special symbol ˑ,
-- then removing remaining ː not after ɕӂ and left paren; do
-- various subs repeatedly in case of multiple geminations in a word
-- 1. immediately after the stress
pron = rsub_repeatedly(pron, '(' .. vowels .. stress_accents .. ')ː(' .. vowels .. ')', '%1ˑ%2')
-- 2. remaining geminate n after the stress between vowels
pron = rsub_repeatedly(pron, '(' .. stress_accents .. '.-' .. vowels .. accents .. '?n)ː(' .. vowels .. ')', '%1(ː)%2')
-- 3. remaining ž and n between vowels
pron = rsub_repeatedly(pron, '(' .. vowels .. accents .. '?)ː(' .. vowels .. ')', '%1ˑ%2')
-- 4. ssk (and zsk, already normalized) immediately after the stress
pron = rsub(pron, '(' .. vowels .. stress_accents .. '*s)ː(k)', '%1ˑ%2')
-- 5. eliminate remaining gemination, except for ɕː and ӂː
pron = rsub(pron, '()ː', '%1')
-- 6. convert special gemination symbol ˑ to regular gemination
pron = rsub(pron, 'ˑ', 'ː')
end
-- handle soft and hard signs, assimilative palatalization
-- 1. insert j before i when required
pron = rsub(pron, 'ʹi', 'ʹji')
-- 2. insert glottal stop after hard sign if required
pron = rsub(pron, 'ʺ()', 'ʔ%1')
-- 3. (ь) indicating optional palatalization
pron = rsub(pron, '%(ʹ%)', '⁽ʲ⁾')
-- 4. assimilative palatalization of consonants when followed by
-- front vowels or soft sign
pron = rsub(pron, '()(*)', '%1ʲ%2')
pron = rsub(pron, '()(*)', '%1ʲ%2')
-- 5. remove hard and soft signs
pron = rsub(pron, "", "")
-- reduction of unstressed word-final -я, -е; but special-case
-- unstressed не, же. Final -я always becomes ; final -е may
-- become , , or depending on the part of speech and
-- the preceding consonants/vowels.
pron = rsub(pron, '⁀', 'ə⁀')
pron = rsub(pron, '⁀nʲe⁀', '⁀nʲi⁀')
pron = rsub(pron, '⁀že⁀', '⁀žy⁀')
-- function to fetch the appropriate value for ending and part of
-- speech, handling aliases and defaults and converting 'e' to 'ê'
-- so that the unstressed sound is preserved
function fetch_e_sub(ending)
local sub = fetch_pos_property(i, ending)
if sub == 'e' then
-- add TEMPCFLEX (which will be converted to CFLEX) to preserve
-- the unstressed sound, which will otherwise be converted
-- to ; we do this instead of adding CFLEX directly because
-- we later convert some instances of the resulting 'e' to
-- 'i', and we don't want to do this when the user explicitly
-- wrote a Cyrillic е with a circumflex on it. [NOTE that
-- formerly applied when we added CFLEX directly: DO NOT
-- use ê here directly because it's a single composed char,
-- when we need the e and accent to be separate.]
return 'e' .. TEMPCFLEX
else
return sub
end
end
if new_final_e_code then
-- handle substitutions in two parts, one for vowel+j+e sequences
-- and the other for cons+e sequences
pron = rsub(pron, vowels_c .. '(' .. accents .. '?j)ë⁀', function(v, ac)
local ty = v == 'o' and 'oe' or 've'
return v .. ac .. fetch_e_sub(ty) .. '⁀'
end)
-- consonant may palatalized, geminated or optional-geminated
pron = rsub(pron, '(.)(ʲ?*)⁀', function(ch, mod)
local ty = ch == 'j' and 'je' or
rfind(ch, '') and 'hardsib' or
rfind(ch, '') and 'softsib' or
'softpaired'
return ch ..mod .. fetch_e_sub(ty) .. '⁀'
end)
if final_e_non_pausal then
-- final should become when not followed by pause or
-- end of utterance (in other words, followed by space plus
-- anything but a pause symbol, or followed by tie bar).
pron = rsub(pron, 'e' .. TEMPCFLEX .. '⁀‿', 'i⁀‿')
if i < #word and word ~= '⁀|⁀' then
pron = rsub(pron, 'e' .. TEMPCFLEX .. '⁀$', 'i⁀')
end
end
-- now convert TEMPCFLEX to CFLEX; we use TEMPCFLEX so the previous
-- two regexps won't affect cases where the user explicitly wrote
-- a circumflex
pron = rsub(pron, TEMPCFLEX, CFLEX)
else
-- Do the old way, which mostly converts final -е to schwa, but
-- has highly broken retraction code for vowel + + е (but
-- not with accent on vowel!) before it that causes final -е in
-- this circumstance to become , and a special hack for кое-.
pron = rsub(pron, vowels_c .. '(*)', '%1%2ɛ')
pron = rsub(pron, '⁀ko(' .. stress_accents .. ')jë⁀', '⁀ko%1ji⁀')
pron = rsub(pron, '⁀', 'ə⁀')
end
-- retraction of е and и after цшж
pron = rsub(pron, '(*)()', function(a, b)
return a .. retracting end)
--syllabify, inserting @ at syllable boundaries
--1. insert @ after each vowel
pron = rsub(pron, '(' .. vowels .. accents .. '?)', '%1@')
--2. eliminate word-final @
pron = rsub(pron, '@+⁀$', '⁀')
--3. move @ forward directly before any ‿⁀, as long as at least
-- one consonant follows that; we will move it across ‿⁀ later
pron = rsub(pron, '@(*)(+)', '%1@%2')
--4. in a consonant cluster, move @ forward so it's before the
-- last consonant
pron = rsub(pron, '@(*)(ʲ?*‿?)', '%1@%2')
--5. move @ backward if in the middle of a "permanent onset" cluster,
-- e.g. sk, str, that comes before a vowel, putting the @ before
-- the permanent onset cluster
pron = rsub(pron, '(?)(_*)()(_*)@()(ʲ?**)', function(a, aund, b, bund, c, d)
if perm_syl_onset or c == 'j' and rfind(b, '') then
return '@' .. a .. aund .. b .. bund .. c .. d
elseif perm_syl_onset then
return a .. aund .. '@' .. b .. bund .. c .. d
end end)
--6. if / is present (explicit syllable boundary), remove any @
-- (automatic boundary) and convert / to @
if rfind(pron, '/') then
pron = rsub(pron, '+', function(x)
if rfind(x, '/') then
x = rsub(x, '@', '')
x = rsub(x, '/', '@')
end
return x
end)
end
--7. remove @ followed by a final consonant cluster
pron = rsub(pron, '@(+⁀)$', '%1')
--8. remove @ preceded by an initial consonant cluster (should only
-- happen when / is inserted by user or in цз, чж sequences)
pron = rsub(pron, '^(⁀+)@', '%1')
--9. make sure @ isn't directly before linking ‿⁀
pron = rsub(pron, '@(+)', '%1@')
-- handle word-initial unstressed o and a; note, vowels always
-- followed by at least one char because of word-final ⁀
-- do after syllabification because syllabification doesn't know
-- about ɐ as a vowel
pron = rsub(pron, '^⁀()', '⁀ɐ%1')
--split by syllable
local syllable = rsplit(pron, '@', true)
--create set of 1-based syllable indexes of stressed syllables
--(acute, grave, circumflex)
local stress = {}
for j = 1, #syllable do
if rfind(syllable, stress_accents) then
stress = "real"
elseif rfind(syllable, CFLEX) then
stress = "cflex"
end
end
-- iterate syllable by syllable to handle stress marks, vowel allophony
local syl_conv = {}
for j = 1, #syllable do
local syl = syllable
local alnum
--vowel allophony
if stress then
-- convert acute/grave/circumflex accent to appropriate
-- IPA marker of primary/secondary/unmarked stress
alnum = 1
syl = rsub(syl, '(.*)́', 'ˈ%1')
syl = rsub(syl, '(.*)̀', 'ˌ%1')
syl = rsub(syl, CFLEX, '')
elseif stress == "real" then
-- special-casing written а immediately before the stress,
-- but only for primary/secondary stress, not circumflex
alnum = 2
else
alnum = 3
end
syl = rsub(syl, vowels_c, function(a)
if a ~= '' then
return allophones
end end)
syl_conv = syl
end
pron = table.concat(syl_conv, "")
-- Optional (j) before ɪ, which is always unstressed
pron = rsub(pron, "⁀jɪ", "⁀(j)ɪ")
pron = rsub(pron, '()jɪ', "%1(j)ɪ")
--consonant assimilative palatalization of tn/dn/sn/zn, depending on
--whether precedes
pron = rsub(pron, '(?)(*)(*nʲ)', function(a, b, c)
if a == '' then
return a .. b .. 'ʲ' .. c
else
return a .. b .. '⁽ʲ⁾' .. c
end end)
--consonant assimilative palatalization of st/zd, depending on
--whether precedes
pron = rsub(pron, '(?)(?)(*ʲ)', function(a, b, c)
if a == '' then
return a .. b .. 'ʲ' .. c
else
return a .. b .. '⁽ʲ⁾' .. c
end end)
--general consonant assimilative palatalization
pron = rsub_repeatedly(pron, '()(*)(ʲ)', function(a, b, c)
if cons_assim_palatal then
return a .. 'ʲ' .. b .. c
elseif cons_assim_palatal then
return a .. '⁽ʲ⁾' .. b .. c
else
return a .. b .. c
end end)
-- further assimilation before alveolopalatals
pron = rsub(pron, 'n(*)()', 'nʲ%1%2')
-- optional palatal assimilation of вп, вб only word-initially
pron = rsub(pron, '⁀(*)(*ʲ)', '⁀%1⁽ʲ⁾%2')
-- optional palatal assimilation of бв but not in обв-
pron = rsub(pron, 'b(*vʲ)', 'b⁽ʲ⁾%1')
if rfind(word, '⁀o' .. accents .. '?bv') then
-- ə in case of a word with a preceding preposition
pron = rsub(pron, '⁀(**)b⁽ʲ⁾(*vʲ)', '⁀%1b%2')
end
-- Word-final -лся (normally in past verb forms) should have optional
-- palatalization. Need to rewrite as -лсьа to defeat this.
-- FIXME: Should we move this to phonetic_subs?
if rfind(word, 'ls⁀') then
pron = rsub(pron, 'lsʲə⁀', 'ls⁽ʲ⁾ə⁀')
end
word = pron
end
text = table.concat(word, " ")
if bracket then
text = ''
end
table.insert(debug, text)
-- Front a and u between soft consonants. If between a soft and
-- optionally soft consonant (should only occur in that order, shouldn't
-- ever have a or u preceded by optionally soft consonant),
-- split the result into two. We only split into two even if there
-- happen to be multiple optionally fronted a's and u's to avoid
-- excessive numbers of possibilities (and it simplifies the code).
-- 1. First, temporarily add soft symbol to inherently soft consonants.
text = rsub(text, '()', '%1ʲ')
-- 2. Handle case of between two soft consonants
text = rsub_repeatedly(text, '(ʲ*)()(?.ʲ)', function(a, b, c)
return a .. fronting .. c end)
-- 3. Handle between soft consonant and optional j, which is still fronted
text = rsub_repeatedly(text, '(ʲ*)()(?%(jʲ%))', function(a, b, c)
return a .. fronting .. c end)
-- 4. Handle case of between soft and optionally soft consonant
if rfind(text, 'ʲ*?.⁽ʲ⁾') then
local opt_hard = rsub(text, '(ʲ*)()(?.)⁽ʲ⁾', '%1%2%3')
local opt_soft = rsub(text, '(ʲ*)()(?.)⁽ʲ⁾', function(a, b, c)
return a .. fronting .. c .. 'ʲ' end)
text = opt_hard .. ', ' .. opt_soft
end
-- 5. Undo addition of soft symbol to inherently soft consonants.
text = rsub(text, '()ʲ', '%1')
-- convert special symbols to IPA
text = rsub(text, 'ʲ', translit_conv_j)
text = rsub(text, '', translit_conv)
-- Assimilation involving hiatus of ɐ and ə
text = rsub(text, 'ə(*)', 'ɐ%1ɐ')
-- eliminate ⁀ symbol at word boundaries
-- eliminate _ symbol that prevents assimilations
text = rsub(text, '', '')
if test_new_ru_pron_module then
if new_module_result ~= text then
--error(text .. " || " .. new_module_result)
track("different-pron")
else
track("same-pron")
end
end
-- return text
table.insert(debug, text)
return table.concat(debug, ':')
end
return export
-- For Vim, so we get 4-space tabs
-- vim: set ts=4 sw=4 noet: