local export = {}
local m_IPA = require("Module:IPA")
local m_hyph = require("Module:fi-hyphenation")
local m_str_utils = require("Module:string utilities")
local find = m_str_utils.find
local gmatch = m_str_utils.gmatch
local gsub = m_str_utils.gsub
local len = m_str_utils.len
local lower = m_str_utils.lower
local match = m_str_utils.match
local sub = m_str_utils.sub
local U = m_str_utils.char
local langcode = "fi"
local nonsyllabic = U(0x32F) -- inverted breve below
local unreleased = U(0x31A)
local nasalized = U(0x303)
local long = "ː"
local letters_phonemes = {
["a"] = "ɑ",
["ä"] = "æ",
["ö"] = "ø",
["å"] = "o",
["g"] = "ɡ",
["q"] = "k",
["v"] = "ʋ",
["š"] = "ʃ",
["ž"] = "ʒ",
["x"] = "ks",
["zz"] = "ts",
["ng"] = "ŋː",
["nk"] = "ŋk",
["nkk"] = "ŋkː",
["qu"] = "kʋ",
["*"] = "ˣ",
["'"] = ".",
}
local vowels = "ɑeiouyæø"
local vowel = "[" .. vowels .. "]"
local consonants = "kptɡgbdfʔsnmŋlrhʋʃʒrjçɦx"
local consonant = "[" .. consonants .. "]"
local diacritics = "̝̞̠̪"
local diacritic = "[" .. diacritics .. "]"
local spelled_consonants = "cšvwxzž"
local spelled_consonant = "[" .. consonants .. spelled_consonants .. "]"
local spelled_vowels = "aäö"
local spelled_vowel = "[" .. vowels .. spelled_vowels .. "]"
local tertiary = "ˌ" -- "tertiary stress", a weaker secondary stress (either rhythmic or in some compound words). is there a better way to represent this?
export.tertiary = tertiary
local stress_indicator = "[ ˈˌ" .. tertiary .. "/+-]"
export.stress_indicator = stress_indicator
local plosives = "kptbdɡ"
local wordlet_sep = "[/-]"
local stress_p = "[ˈˌ" .. tertiary .. "]"
local stress_s = "[ˌ" .. tertiary .. "]"
local stress_pd = "[ˈˌ" .. tertiary .. "%.]"
local stress_sd = "[ˌ" .. tertiary .. "%.]"
local replacements_narrow = {
["ɑ"] = "ɑ̝",
["e"] = "e̞",
["ø"] = "ø̞",
["o"] = "o̞",
["t"] = "t̪",
["s"] = "s̠"
}
-- This adds letters_phonemes["e"] = "e", letters_phonemes["i"] = "i", etc.
for letter in gmatch("eiouydhfjklmnprstu", ".") do
letters_phonemes[letter] = letter
end
--[[ This regex finds the diphthongs in the IPA transcription,
so that the nonsyllabic diacritic can be added. ]]
-- /_i/ diphthongs can appear in any syllable
local diphthongs_i = {
"[ɑeouyæø]i"
}
-- /_U/ diphthongs can appear in the initial syllable or later open syllables (no consonantal coda)
local diphthongs_u = {
"[ɑoei]u",
"[eiæø]y",
}
-- rising diphthongs can only appear in the initial syllable (of a word, compound word part, etc.)
local diphthongs_rising = {
"uo",
"ie",
"yø",
}
local function apply_post_fixes(p)
-- initial <gn> is /gn/
p = gsub(p, "ˈŋn", "ˈɡn")
-- ŋ is short before consonant (by default)
p = gsub(p, "ŋ"..long.."("..consonant..")", "ŋ%1")
-- dissimilation of vowels by sandhi
p = gsub(p, "("..vowel..diacritic.."*)(["..long..nonsyllabic.."]?)("..stress_s..")%1", "%1%2%3(ʔ)%1")
-- C1(.):C2 -> C1(.)(:)C2
p = gsub(p, "("..consonant..diacritic.."*)"..long.."("..consonant..diacritic.."*)", "%1("..long..")%2")
p = gsub(p, "("..consonant..diacritic.."*)%.%1("..consonant..diacritic.."*)", "%1.(%1)%2")
p = gsub(p, "("..consonant..diacritic.."*)%(%.?%)%1("..consonant..diacritic.."*)", "%1(.%1)%2")
return p
end
local function apply_post_fixes_narrow(p)
-- C1:+C2 -> C1+(C1)C2
p = gsub(p, "("..consonant..diacritic.."*)"..long..tertiary.."("..consonant..diacritic.."*)", "%1"..tertiary.."(%1)%2")
-- C1+C1C2 -> C1+(C1)C2
-- but C1C2+C2C3 --> C1(C2)+C2C3
p = gsub(p, "("..consonant..diacritic..")("..consonant..diacritic.."*)"..tertiary.."(%.?)%2("..consonant..diacritic.."*)", "%1(%2)"..tertiary.."%2%3%4")
p = gsub(p, "("..consonant..diacritic.."*)"..tertiary.."(%.?)%1("..consonant..diacritic.."*)", "%1"..tertiary.."(%1)%2%3")
p = gsub(p, "("..consonant..diacritic.."*)"..tertiary.."%(?%.?%)%1("..consonant..diacritic.."*)", "%1"..tertiary.."(.%1)%2")
-- t is alveolar in /ts/ and /st/
p = gsub(p, "t̪("..stress_pd.."?%s*)s̠", "t%1s̠")
p = gsub(p, "s̠("..stress_pd.."?%s*)t̪", "s̠%1t")
-- n is dental in /nt/
p = gsub(p, "n("..stress_pd.."?%s*)t̪", "n̪%1t̪")
p = gsub(p, "t̪("..stress_pd.."?%s*)n", "t̪%1n̪")
-- l allophone
p = gsub(p, "l("..stress_pd.."?%s*)t̪", "l̪%1t̪")
p = gsub(p, "t̪("..stress_pd.."?%s*)l", "t̪%1l̪")
-- long j, v after i, u diphthong
p = gsub(p, "(i"..nonsyllabic..")j("..vowel..")", "%1j("..long..")%2")
-- /ʋ/ after /u/ usually realized as /w/ (see Suomi, Toivanen and Ylitalo 2008, p. )
p = gsub(p, "(u"..nonsyllabic..")ʋ("..vowel..")", "%1w("..long..")%2")
-- cleanup
p = gsub(p, "("..stress_s..")%.", "%1")
p = gsub(p, "("..stress_pd..")"..stress_s, "%1")
-- tautosyllabic nasals nasalize vowels between them (see Suomi, Toivanen and Ylitalo 2008, p. 22)
--p = gsub(p, "([mnŋ]"..long.."?)("..vowel..")("..diacritic.."*)([mnŋ])(.?)", function (n0, nv, nvd, n1, anchor)
-- -- this cannot be simplified to "(.?)" => "([^" .. vowels .. "]?)", otherwise a vowel after would match
-- if not find(anchor, vowel) then
-- return n0 .. nv .. nasalized .. nvd .. n1 .. anchor
-- end
--end)
-- sandhi: nm > mm, np > mp, nb > mb, nk > ŋk, ng > ŋg
p = gsub(p, "nm", "m" .. long)
p = gsub(p, "n("..stress_pd.."?%s*)([ɡk])", "ŋ%1%2")
p = gsub(p, "n("..stress_pd.."?%s*)([mpb])", "m%1%2")
p = gsub(p, "[nm]("..stress_pd.."?%s*)([f])", "ɱ%1%2")
-- handle potentially long consonants over secondary stresses
p = gsub(p, "("..consonant..diacritic.."*)%("..long.."%)("..stress_s..")", "%2%1("..long..")")
p = gsub(p, "("..stress_s..")("..consonant..diacritic.."*)%("..long.."%)", "(%2)%1%2")
p = gsub(p, "%(([kptbdɡ])%)("..stress_s..")%1", "(%1"..unreleased..")%2%1")
p = gsub(p, "(ŋ"..diacritic.."*)"..tertiary.."ɡ", "%1"..tertiary.."ŋ")
-- [k] allophone before front vowels (see Suomi, Toivanen and Ylitalo 2008, p. 27)
p = gsub(p, "k([eiyæø])", "k̟%1")
return p
end
-- -1 = degenerate
-- 0 = light
-- 1 = heavy with coda
-- 2 = heavy without coda (long vowel/diphthong)
-- 99 = stressed
local class_stressed_syllable = 99
local function classify_syllable(syllable)
if match(syllable, "^" .. stress_indicator) then
return class_stressed_syllable
end
local cleaned = gsub(syllable, "%(?%*?%)?$", "")
if match(cleaned, spelled_vowel .. spelled_vowel .. "$") then
return 2
elseif match(cleaned, spelled_consonant .. "$") then
return match(cleaned, spelled_vowel) and 1 or -1
elseif match(cleaned, spelled_vowel) then
return 0
end
return -1
end
local function has_hiatus(left, right)
return match(left, spelled_vowel .. "$") and match(right, "^%(?%.?%)?" .. spelled_vowel)
end
local trad_lookahead = 2 -- how many unstressed syllables at most in a single unit, thus max consecutive unstressed syllables
local function add_rhythmic_stress_ltr(hyph, classes)
-- traditional syllable stress assigning algorithm (e.g. in (V)ISK)
-- assign from left to right, but shift to later syllables if a light syllable is followed by a heavy syllable
local add_stress = {}
local n = #classes
local prev_stress
for index, class in ipairs(classes) do
local stressed = class == class_stressed_syllable
if not stressed and not prev_stress and index < n then
-- shift stress if current syllable light and a heavy syllable occurs later (except as the last syllable)
local has_later_heavy_syllable
local lookahead_end = math.min(index + trad_lookahead, n - 1)
for i = index + 1, lookahead_end do
if classes[i] > 0 then
has_later_heavy_syllable = true
break
end
end
stressed = index == n - 1 or (class > 0) or not has_later_heavy_syllable
end
-- check if next syllable already stressed, and do not stress this syllable if so
if stressed and index < n then
stressed = stressed and classes[index + 1] ~= class_stressed_syllable
end
if index > 1 and stressed then
add_stress[index] = true
end
prev_stress = stressed
end
return add_stress
end
local function add_rhythmic_stress_rtl(hyph, classes)
-- based on Karvonen (2005) "Word prosody in Finnish"
-- run algorithm for assigning stresses:
-- (1) analyze each segment started by an already stressed syllable
-- as a separate prosodic word
-- (2) assign from *right to left*
-- (3) final syllable is never stressed (not by this algorithm)
-- (4) stress may shift back by a single (unstressed) syllable
-- to a better candidate:
-- (C)VV is better than (C)V
-- (C)VC is better than (C)V and (C)VV
-- (5) if the final syllable in a prosodic word begins with a vowel
-- and the previous syllable ends in a vowel, skip it
local add_stress = {}
local foot = 0
local word_final = true
local index = #classes
while index > 0 do
local stressed = classes[index] == class_stressed_syllable
if stressed then
-- reset
foot = 0
word_final = true
elseif word_final and index > 1 and has_hiatus(hyph[index - 1], hyph[index]) then
-- hiatus break at the end of a prosodic word: do nothing
elseif foot == 0 then
foot = foot + 1
-- try to assign a stress.
-- is the previous syllable stressed?
elseif index > 1 and classes[index - 1] == class_stressed_syllable then
-- no two stresses in a row, so do nothing.
else
-- we will assign a stress somewhere.
local class = classes[index]
-- is this syllable preceded by an unstressed syllable
-- with more priority?
if index > 2 and classes[index - 2] ~= class_stressed_syllable and
classes[index - 1] > classes[index] then
-- shift back by one syllable
index = index - 1
end
-- assign stress, and continue.
add_stress[index] = true
foot = 0
end
index = index - 1
word_final = false
end
return add_stress
end
local function add_rhythmic_stress_collect(hyph, classes, add_stress)
-- add stress marks
local rhythmic_stress = tertiary
local collected = {}
for index = 1, #hyph do
if add_stress[index] and classes[index] < class_stressed_syllable then
collected[index] = rhythmic_stress .. gsub(hyph[index], "^%.", "")
else
collected[index] = hyph[index]
end
end
return table.concat(collected)
end
-- applied *before* IPA conversion
-- returns one result, or possibly two results
function export.add_rhythmic_stress(word, assign_rtl)
-- keep_sep_symbols = true
local hyph = m_hyph.generate_hyphenation(word, true)
local final_heavy = false
local collected
local add_rhythmic_stress_impl = assign_rtl and add_rhythmic_stress_rtl or add_rhythmic_stress_ltr
-- find stressed syllables and classify all syllables
local classes = {}
for index, syllable in ipairs(hyph) do
if index == 1 then
classes[index] = class_stressed_syllable
else
classes[index] = classify_syllable(syllable)
end
end
local add_stress
-- final heavy syllable may be stressed if preceded by an unstressed
-- light syllable (in words with at least 3 syllables)
final_heavy = #classes >= 3 and classes[#classes] > 0 and classes[#classes - 1] == 0
add_stress = add_rhythmic_stress_impl(hyph, classes)
local res1 = add_rhythmic_stress_collect(hyph, classes, add_stress)
local res2
if final_heavy then
-- make final syllable stressed
local actual_final = classes[#classes]
classes[#classes] = class_stressed_syllable
add_stress = add_rhythmic_stress_impl(hyph, classes)
-- so that add_rhythmic_stress_collect actually adds it
classes[#classes] = actual_final
add_stress[#classes] = true
res2 = add_rhythmic_stress_collect(hyph, classes, add_stress)
end
return res1, res2
end
local function handle_diphthongs(IPA, strict_initial)
-- Add nonsyllabic diacritic after last vowel of diphthong.
for _, diphthong_regex in pairs(diphthongs_i) do
IPA = gsub(IPA, diphthong_regex, "%0" .. nonsyllabic)
end
local only_initial = (strict_initial and ("^[^" .. vowels .. "]*")
or (stress_indicator .. "[^" .. vowels .. "]*"))
for _, diphthong_regex in pairs(diphthongs_rising) do
-- initial syllables
IPA = gsub(IPA, only_initial .. diphthong_regex, "%0" .. nonsyllabic)
end
for _, diphthong_regex in pairs(diphthongs_u) do
-- initial syllables
IPA = gsub(IPA, only_initial .. diphthong_regex, "%0" .. nonsyllabic)
local open_noninitial =
function(diphthong, after)
if find(after, "^" .. consonant .. diacritic .. "*" .. vowel) then
-- consonant after diphthong
-- must be followed by vowel so that it's part of the
-- following syllable, else it's in this syllable
-- and thus this syllabie is closed
return diphthong .. nonsyllabic .. after
elseif find(after, "^" .. consonant) then
-- consonant after diphthong
-- must be in this syllable
return diphthong .. after
end
-- no consonant after diphthong => open
return diphthong .. nonsyllabic .. after
end
-- open non-initial syllables
IPA = gsub(IPA, "(" .. diphthong_regex .. ")([^" .. nonsyllabic .. "].+)", open_noninitial)
IPA = gsub(IPA, "(" .. diphthong_regex .. ")($)", open_noninitial)
end
return IPA
end
local function IPA_wordlet(term, is_narrow, has_initial)
local rest = term
local phonemes = {}
while len(rest) > 0 do
-- Find the longest string of letters that matches a recognised sequence in the list
local longestmatch = ""
for letter, phoneme in pairs(letters_phonemes) do
if sub(rest, 1, len(letter)) == letter and len(letter) > len(longestmatch) then
longestmatch = letter
end
end
-- Convert the string to IPA
if len(longestmatch) > 0 then
table.insert(phonemes, letters_phonemes[longestmatch])
rest = sub(rest, len(longestmatch) + 1)
else
table.insert(phonemes, sub(rest, 1, 1))
rest = sub(rest, 2)
end
end
local result = table.concat(phonemes)
if is_narrow then
-- articulation of h (Suomi, Toivanen & Ylitalo 2008, p. 28)
result = gsub(result, "(.?)h(.?)",
function (before, after)
local h
if after ~= "" and after ~= "h" then
if before ~= "" and vowels:find(before) then
if consonants:find(after) then
-- vihma, yhtiö
if before == "i" or before == "y" then
h = "ç"
-- mahti, kohme, tuhka
elseif before == "ɑ" or before == "o" or before == "u" then
h = "x"
end
-- maha
elseif vowels:find(after) then
h = "ɦ"
end
end
end
if h then
return before .. h .. after
end
end)
-- double letter replacement and diphthongs must be handled earlier here
result = gsub(result, "(%a)%1", "%1" .. long)
if has_initial then
result = handle_diphthongs(result, true)
end
for letter, phoneme in pairs(replacements_narrow) do
result = gsub(result, letter, phoneme)
end
end
return result
end
function export.IPA_term(term, is_narrow)
term = lower(term)
local notinitial = {} -- true if the component is not an initial component
local hyphenstress = "ˌ" -- secondary by default
local is_prefix = false
local is_suffix = false
if find(term, "%/") then
hyphenstress = tertiary -- tertiary if we have slashes
end
local found
term, found = gsub(term, "^%-+", "")
is_suffix = found > 0
term, found = gsub(term, "%-+$", "")
is_prefix = found > 0
if is_narrow then
term = export.add_rhythmic_stress(term)
end
-- scan for wordlets
local wordlet_start = 1
local wordlet_IPAs = {}
while true do
local wordlet_sep_at, _, wordlet_sep = find(term, "(" .. wordlet_sep .. ")", wordlet_start)
local wordlet_end = wordlet_sep_at and (wordlet_sep_at - 1)
local part = sub(term, wordlet_start, wordlet_end)
local stress
if wordlet_start == 1 then
stress = "ˈ" -- primary
elseif wordlet_sep == "/" then
stress = "ˌ" -- secondary
else
stress = hyphenstress
end
table.insert(wordlet_IPAs, stress .. IPA_wordlet(part, is_narrow, true))
if wordlet_sep_at == nil then
break
else
wordlet_start = wordlet_sep_at + 1
end
end
IPA = table.concat(wordlet_IPAs, "")
if is_narrow then
-- handle * in narrow transcription
IPA = gsub(IPA, "ˣ(%)?%s*"..stress_p.."?)((.?)" .. diacritic .. "*)",
function (post, after, potential_consonant)
if potential_consonant == "" then
if find(post, "^%)") then
return "ʔ" .. post .. after
else
return post .. "(ʔ)" .. after
end
elseif consonants:find(potential_consonant) then
if #post > 0 then
local amark = ""
if plosives:find(sub(after, 1, 1)) then
amark = unreleased
end
return after .. amark .. post .. after
else
return post .. after .. long
end
else
return post .. "ʔ" .. after
end
end)
else
-- Replace double letters (vowels or consonants) with single letter plus length sign.
IPA = gsub(IPA, "(%a)%1", "%1" .. long)
IPA = handle_diphthongs(IPA, false)
end
IPA = apply_post_fixes(IPA)
if is_narrow then
IPA = apply_post_fixes_narrow(IPA)
end
if is_prefix then
IPA = IPA .. "-"
end
if is_suffix then
IPA = "-" .. IPA
end
return IPA
end
function export.IPA(term)
if type(term) == "table" then
term = term:getParent().args[1]
end
local title = mw.title.getCurrentTitle().text
if not term then
term = title
elseif term == "*" then
term = title .. "*"
end
local no_count = match(term, " ")
IPA_narrow = export.IPA_term(term, true)
IPA = export.IPA_term(term, false)
return m_IPA.format_IPA_full {
lang = require("Module:languages").getByCode(langcode),
items = {{pron = "/" .. IPA .. "/"}, {pron = "[" .. IPA_narrow .. "]"}},
no_count = no_count,
}
end
return export
