Module:IPA/data

local data = {}

--[=[ This should list the language codes of all languages that have a pronunciation page in the appendix of the form Appendix:LANG pronunciation, e.g. Appendix:Russian pronunciation. For these languages, the text "key" next to the generated pronunciation links to such pages; for other languages, it links to the "LANG phonology" page in Wikipedia (which may or may not exist). Module:IPA is responsible for this linking; see format_IPA_full. ]=] local langs_with_infopages = { "acw", "ady", "ang", "arc", "ba", "bg", "bo", "ca", "cho", "cmn", "cs", "cy", "da", "de", "dsb", "dz", "egl", "egy", "el", "en", "enm", "eo", "es", "fa", "fi", "fo", "fr", "fy", "ga", "gd", "gem-pro", "got", "he", "hi", "hu", "hy", "ii", "is", "it", "iu", "ja", "jbo", "ka", "kls", "ko", "kw", "la", "lb", "liv", "lt", "lv", "mdf", "mfe", "mic", "mk", "ms", "mt", "mul", "my", "nan", "nci", "nl", "nn", "no", "nov", "nv", "pjt", "pl", "ps", "pt", "ro", "ru", "scn", "sco", "sga", "sh", "sl", "sq", "sv", "sw", "syc", "th", "tl", "tr", "ug", "uk", "vi", "vo", "wlm", "yi", "yue", }

data.langs_with_infopages = {}

-- Convert the list in `langs_with_infopages` to a set. for _, langcode in ipairs(langs_with_infopages) do	data.langs_with_infopages[langcode] = true end

--[=[ This should list the diphthongs of a language (in the form of Lua patterns), provided they do *NOT* contain semivowel symbols such as /j w ɰ ɥ/ or vowels with nonsyllabic diacritics such as /i̯ u̯/. For example, list /au/ or /aʊ/, but do not list /aw/ or /au̯/. The data in this table is used to count the number of syllables in a word. Module:syllables automatically knows how to correctly handle semivowel symbols and nonsyllabic diacritics.

Any language listed here will automatically have categories of the form "LANG #-syllable words" generated. In addition, any language listed below under `langs_to_generate_syllable_count_categories` will also have such categories generated.

NOTE: There are some additional languages that have these categories. For example: ]=] data.diphthongs = { ["cs"] = { -- Czech phonology "[aeo]u", },	["de"] = { "a[ɪʊ]", "ɔ[ʏɪ]", },	["en"] = { -- from Appendix:English pronunciation mostly, but /ʌɪ/ is from the OED "[aɑeɛoɔʌ][ɪi]", "[ɑɒæo]e", "[əɐ]ʉ", "[aɒəoɔʌ]ʊ", "æo", "[ɛeɪiɔʊʉ]ə",	-- /iə/ is a diphthong in NZE, but a disyllabic sequence in GA. -- /ɪə/ is both a disyllabic sequence and a diphthong in old-fashioned RP. "[aʌ][ʊɪ]ə",	-- May be a disyllabic sequence in some or all dialects? },	["es"] = { -- Module:es-pronunc generates both phonemic and phonetic variants. The former is first -- and is thus what is used. If the following can't be made reliable using this, we need to		-- modify Module:IPA to use the phonetic variant when computing diphthongs for Spanish; -- in that case, no special entry for Spanish is needed in this table, so it would be added -- to the following table. "[aeou]i", "[aeiu]u", },	["grc"] = { "[aeyo]i", "[ae]u", "[ɛɔa]ː[iu]", },	["is"] = {			-- Icelandic phonology "[aeø][iɪy]",	-- Wikipedia is oddly specific about the second element: ei and ai, but øɪ. "[ao]u", },	["it"] = { "[aeɛoɔu]i", "[aeɛioɔ]u", },	["la"] = { "[eaou]i", "[eao]u", "[ao]e", },	["lb"] = { "[iu]ə", "[ɜoæɑ]ɪ", "[əæɑ]ʊ", }, }
 * Thai words have these categories added by Module:th-pron.

--[=[ This should list any languages for which categories of the form "LANG #-syllable words", e.g. Category:Russian 3-syllable words, should be generated. Do not list languages here if they have an entry above under `data.diphthongs`; such languages are automatically added to this list. ]=] local langs_to_generate_syllable_count_categories = { "ar",	-- Arabic has diphthongs, but they are transcribed -- with semivowel symbols. "ary",	-- Moroccan Arabic has diphthongs, but they are transcribed -- with semivowel symbols. "ca",	-- Catalan has diphthongs, but they are generally transcribed using -- /w/ and /j/, so do not need to be listed (see Catalan language.	"fi",	-- Finnish has diphthongs, but they are now automatically transcribed with			-- the nonsyllabic diacritic	"fr",	-- French has diphthongs, but they are transcribed			-- with semivowel symbols: French phonology.	"ka",   "kmr",	"ku",    "mt",	-- Maltese has diphthongs, but they are transcribed			-- with semivowel symbols.	"ru",	-- No diphthongs, properly speaking; sequences of a vowel and /j/ though.	"sk",	-- Slovak has rising diphthongs, /i̯e, i̯a, i̯u, u̯o/, which are probably always spelled with the nonsyllabic diacritic, so do not need to be listed.	"sl",	-- No diphthongs, properly speaking; sequences of a vowel, /j/ and /w/ though	"sq",	-- Albanian language doesn't mention anything about diphthongs.	"ug",	-- No diphthongs. }

data.langs_to_generate_syllable_count_categories = {}

-- Convert the list in `langs_to_generate_syllable_count_categories` to a set. for _, langcode in ipairs(langs_to_generate_syllable_count_categories) do	data.langs_to_generate_syllable_count_categories[langcode] = true end -- Also add languages listed under `data.diphthongs`. for langcode, _ in pairs(data.diphthongs) do	data.langs_to_generate_syllable_count_categories[langcode] = true end

-- Non-standard or obsolete IPA symbols. data.nonstandard = { --	The following symbols consist of more than one character,			so we can't put them in the line below. "ɑ̢", "d̂", "t̂", "n̂", "l̂", "k̫", "ɔ̗", "ɔ̖", "[ʦʣʧʤʨʥ?ƍσƺƪƻƾƞᶀᶁᶂᶃᶄᶅᶆᶈᶇᶉᶊᶋƫᶌᶍᶎʓʆλƛłščžǰǧǯẋᵻᵿⱻʚ" .. "ɷωıȹȸ∅ØƥƭƈƙʠʇʗʖʞɩɼȣяɿʅʮʯᴀᴀᴇGRŒQȡȶȵȴKPT]" }

-- See valid IPA characters at Module:IPA/data/symbols.

data.phonemes = {} data.phonemes["dz"] = { "m", "n", "ŋ", "p", "t", "ʈ", "k", "pʰ", "tʰ", "ʈʰ", "kʰ", "t͡s", "t͡ɕ", "t͡sʰ", "t͡ɕʰ", "w", "s", "z", "ɬ", "l", "r", "ɕ", "ʑ", "j", "h", "ɑ", "e", "i", "o", "u", "ɑː", "eː", "ɛː", "iː", "oː", "øː", "uː", "yː", "ɑ˥", "e˥", "i˥", "o˥", "u˥", "ɑː˥", "eː˥", "ɛː˥", "iː˥", "oː˥", "øː˥", "uː˥", "yː˥", "m˥", "n˥", "ŋ˥", "p˥", "k˥", "k̚˥", "w˥", "l˥", "r˥", "ɕ˥", "j˥", ")˥",	"ɑ˩", "e˩", "i˩", "o˩", "u˩",	"ɑː˩", "eː˩", "ɛː˩", "iː˩", "oː˩", "øː˩", "uː˩", "yː˩",	"m˩", "n˩", "ŋ˩", "p˩", "k˩", "k̚˩", "w˩", "l˩", "r˩", "ɕ˩", "j˩", ")˩", ".", ",", "-", } data.phonemes["eo"] = { "a", "b", "d", "d͡ʒ", "e", "f", "h", "i", "j", "k", "l", "m", "n", "o", "p", "r", "s", "t", "t͡s", "t͡ʃ", "u", "v", "w", "x", "z", "ɡ", "ʃ", "ʒ", "ˈ", ".", " ", "-", } data.phonemes["hy"] = { "ɑ", "b", "ɡ", "d", "ɛ", "z", "ɛ", "ə", "tʰ", "ʒ", "i", "l", "χ", "t͡s", "k", "h", "d͡z", "ʁ", "t͡ʃ", "m", "j", "n", "ʃ", "ɔ", "t͡ʃʰ", "p", "d͡ʒ", "r", "s", "v", "t", "ɾ", "t͡sʰ", "v", "pʰ", "kʰ", "ɔ", "f", "ŋɡ", "ŋk", "ŋχ", "u", "ˈ", "ˌ", ".", " ", "ː", } data.phonemes["nl"] = { "m", "n", "ŋ", "p", "b", "t", "d", "k", "ɡ", "f", "v", "s", "z", "ʃ", "ʒ", "x", "ɣ", "ɦ", "ʋ", "l", "j", "r", "ɪ", "ʏ", "ɛ", "ə", "ɔ", "ɑ", "i", "iː", "y", "yː", "u", "uː", "eː", "øː", "oː", "ɛː", "œː", "ɔː", "aː", "ɛi̯", "œy̯", "ɔi̯", "ɑu̯", "ɑi̯", "iu̯", "yu̯", "ui̯", "eːu̯", "oːi̯", "aːi̯", "ˈ", "ˌ", ".", " ", }

return data