vits-models / text /cleaners.py
SayaSS
init
bd603fe
raw
history blame
13.5 kB
""" from https://github.com/keithito/tacotron """
'''
Cleaners are transformations that run over the input text at both training and eval time.
Cleaners can be selected by passing a comma-delimited list of cleaner names as the "cleaners"
hyperparameter. Some cleaners are English-specific. You'll typically want to use:
1. "english_cleaners" for English text
2. "transliteration_cleaners" for non-English text that can be transliterated to ASCII using
the Unidecode library (https://pypi.python.org/pypi/Unidecode)
3. "basic_cleaners" if you do not want to transliterate (in this case, you should also update
the symbols in symbols.py to match your data).
'''
import re
from unidecode import unidecode
import pyopenjtalk
from jamo import h2j, j2hcj
from pypinyin import lazy_pinyin, BOPOMOFO
import jieba, cn2an
# This is a list of Korean classifiers preceded by pure Korean numerals.
_korean_classifiers = 'ꡰ데 ꢌ 개 그루 λ‹’ λŒ€ 두 마리 λͺ¨ λͺ¨κΈˆ λ­‡ 발 발짝 λ°© 번 벌 보루 μ‚΄ 수 술 μ‹œ 쌈 움큼 μ • 짝 채 μ²™ 첩 μΆ• 케레 톨 톡'
# Regular expression matching whitespace:
_whitespace_re = re.compile(r'\s+')
# Regular expression matching Japanese without punctuation marks:
_japanese_characters = re.compile(r'[A-Za-z\d\u3005\u3040-\u30ff\u4e00-\u9fff\uff11-\uff19\uff21-\uff3a\uff41-\uff5a\uff66-\uff9d]')
# Regular expression matching non-Japanese characters or punctuation marks:
_japanese_marks = re.compile(r'[^A-Za-z\d\u3005\u3040-\u30ff\u4e00-\u9fff\uff11-\uff19\uff21-\uff3a\uff41-\uff5a\uff66-\uff9d]')
# List of (regular expression, replacement) pairs for abbreviations:
_abbreviations = [(re.compile('\\b%s\\.' % x[0], re.IGNORECASE), x[1]) for x in [
('mrs', 'misess'),
('mr', 'mister'),
('dr', 'doctor'),
('st', 'saint'),
('co', 'company'),
('jr', 'junior'),
('maj', 'major'),
('gen', 'general'),
('drs', 'doctors'),
('rev', 'reverend'),
('lt', 'lieutenant'),
('hon', 'honorable'),
('sgt', 'sergeant'),
('capt', 'captain'),
('esq', 'esquire'),
('ltd', 'limited'),
('col', 'colonel'),
('ft', 'fort'),
]]
# List of (hangul, hangul divided) pairs:
_hangul_divided = [(re.compile('%s' % x[0]), x[1]) for x in [
('γ„³', 'γ„±γ……'),
('γ„΅', 'γ„΄γ…ˆ'),
('γ„Ά', 'γ„΄γ…Ž'),
('γ„Ί', 'γ„Ήγ„±'),
('γ„»', 'ㄹㅁ'),
('γ„Ό', 'γ„Ήγ…‚'),
('γ„½', 'γ„Ήγ……'),
('γ„Ύ', 'γ„Ήγ…Œ'),
('γ„Ώ', 'ㄹㅍ'),
('γ…€', 'γ„Ήγ…Ž'),
('γ…„', 'γ…‚γ……'),
('γ…˜', 'ㅗㅏ'),
('γ…™', 'ㅗㅐ'),
('γ…š', 'γ…—γ…£'),
('ㅝ', 'γ…œγ…“'),
('γ…ž', 'γ…œγ…”'),
('γ…Ÿ', 'γ…œγ…£'),
('γ…’', 'γ…‘γ…£'),
('γ…‘', 'ㅣㅏ'),
('γ…’', 'ㅣㅐ'),
('γ…•', 'γ…£γ…“'),
('γ…–', 'γ…£γ…”'),
('γ…›', 'γ…£γ…—'),
('γ… ', 'γ…£γ…œ')
]]
# List of (Latin alphabet, hangul) pairs:
_latin_to_hangul = [(re.compile('%s' % x[0], re.IGNORECASE), x[1]) for x in [
('a', '에이'),
('b', 'λΉ„'),
('c', 'μ‹œ'),
('d', 'λ””'),
('e', '이'),
('f', '에프'),
('g', '지'),
('h', 'μ—μ΄μΉ˜'),
('i', '아이'),
('j', '제이'),
('k', '케이'),
('l', 'μ—˜'),
('m', 'μ— '),
('n', 'μ—”'),
('o', '였'),
('p', 'ν”Ό'),
('q', '큐'),
('r', 'μ•„λ₯΄'),
('s', 'μ—μŠ€'),
('t', 'ν‹°'),
('u', '유'),
('v', '브이'),
('w', 'λ”λΈ”μœ '),
('x', 'μ—‘μŠ€'),
('y', '와이'),
('z', '제트')
]]
# List of (Latin alphabet, bopomofo) pairs:
_latin_to_bopomofo = [(re.compile('%s' % x[0], re.IGNORECASE), x[1]) for x in [
('a', 'γ„ŸΛ‰'),
('b', 'ㄅㄧˋ'),
('c', 'ㄙㄧˉ'),
('d', 'ㄉㄧˋ'),
('e', 'ㄧˋ'),
('f', 'γ„ΛŠγ„ˆγ„¨Λ‹'),
('g', 'ㄐㄧˋ'),
('h', 'ㄝˇㄑㄩˋ'),
('i', 'γ„žΛ‹'),
('j', 'γ„γ„ŸΛ‹'),
('k', 'γ„Žγ„ŸΛ‹'),
('l', 'γ„ΛŠγ„›Λ‹'),
('m', 'γ„ΛŠγ„‡γ„¨Λ‹'),
('n', 'γ„£Λ‰'),
('o', 'γ„‘Λ‰'),
('p', 'ㄆㄧˉ'),
('q', 'γ„Žγ„§γ„‘Λ‰'),
('r', 'γ„šΛ‹'),
('s', 'γ„ΛŠγ„™Λ‹'),
('t', 'γ„Šγ„§Λ‹'),
('u', 'ㄧㄑˉ'),
('v', 'ㄨㄧˉ'),
('w', 'γ„‰γ„šΛ‹γ„…γ„¨Λ‹γ„Œγ„§γ„‘Λ‹'),
('x', 'γ„Λ‰γ„Žγ„¨Λ‹γ„™Λ‹'),
('y', 'γ„¨γ„žΛ‹'),
('z', 'γ„—γ„ŸΛ‹')
]]
# List of (bopomofo, romaji) pairs:
_bopomofo_to_romaji = [(re.compile('%s' % x[0], re.IGNORECASE), x[1]) for x in [
('γ„…γ„›', 'p⁼wo'),
('ㄆㄛ', 'pΚ°wo'),
('ㄇㄛ', 'mwo'),
('γ„ˆγ„›', 'fwo'),
('γ„…', 'p⁼'),
('ㄆ', 'pΚ°'),
('ㄇ', 'm'),
('γ„ˆ', 'f'),
('ㄉ', 't⁼'),
('γ„Š', 'tΚ°'),
('γ„‹', 'n'),
('γ„Œ', 'l'),
('ㄍ', 'k⁼'),
('γ„Ž', 'kΚ°'),
('ㄏ', 'h'),
('ㄐ', 'ʧ⁼'),
('γ„‘', 'Κ§Κ°'),
('γ„’', 'Κƒ'),
('γ„“', 'Κ¦`⁼'),
('γ„”', 'Κ¦`Κ°'),
('γ„•', 's`'),
('γ„–', 'ΙΉ`'),
('γ„—', 'ʦ⁼'),
('γ„˜', 'Κ¦Κ°'),
('γ„™', 's'),
('γ„š', 'a'),
('γ„›', 'o'),
('γ„œ', 'Ι™'),
('ㄝ', 'e'),
('γ„ž', 'ai'),
('γ„Ÿ', 'ei'),
('γ„ ', 'au'),
('γ„‘', 'ou'),
('ㄧㄒ', 'yeNN'),
('γ„’', 'aNN'),
('ㄧㄣ', 'iNN'),
('γ„£', 'Ι™NN'),
('γ„€', 'aNg'),
('ㄧγ„₯', 'iNg'),
('ㄨγ„₯', 'uNg'),
('γ„©γ„₯', 'yuNg'),
('γ„₯', 'Ι™Ng'),
('ㄦ', 'Ι™Ι»'),
('ㄧ', 'i'),
('ㄨ', 'u'),
('γ„©', 'Ι₯'),
('Λ‰', 'β†’'),
('ˊ', '↑'),
('Λ‡', '↓↑'),
('Λ‹', '↓'),
('Λ™', ''),
(',', ','),
('。', '.'),
('!', '!'),
('?', '?'),
('β€”', '-')
]]
def expand_abbreviations(text):
for regex, replacement in _abbreviations:
text = re.sub(regex, replacement, text)
return text
def lowercase(text):
return text.lower()
def collapse_whitespace(text):
return re.sub(_whitespace_re, ' ', text)
def convert_to_ascii(text):
return unidecode(text)
def japanese_to_romaji_with_accent(text):
'''Reference https://r9y9.github.io/ttslearn/latest/notebooks/ch10_Recipe-Tacotron.html'''
sentences = re.split(_japanese_marks, text)
marks = re.findall(_japanese_marks, text)
text = ''
for i, sentence in enumerate(sentences):
if re.match(_japanese_characters, sentence):
if text!='':
text+=' '
labels = pyopenjtalk.extract_fullcontext(sentence)
for n, label in enumerate(labels):
phoneme = re.search(r'\-([^\+]*)\+', label).group(1)
if phoneme not in ['sil','pau']:
text += phoneme.replace('ch','Κ§').replace('sh','Κƒ').replace('cl','Q')
else:
continue
n_moras = int(re.search(r'/F:(\d+)_', label).group(1))
a1 = int(re.search(r"/A:(\-?[0-9]+)\+", label).group(1))
a2 = int(re.search(r"\+(\d+)\+", label).group(1))
a3 = int(re.search(r"\+(\d+)/", label).group(1))
if re.search(r'\-([^\+]*)\+', labels[n + 1]).group(1) in ['sil','pau']:
a2_next=-1
else:
a2_next = int(re.search(r"\+(\d+)\+", labels[n + 1]).group(1))
# Accent phrase boundary
if a3 == 1 and a2_next == 1:
text += ' '
# Falling
elif a1 == 0 and a2_next == a2 + 1 and a2 != n_moras:
text += '↓'
# Rising
elif a2 == 1 and a2_next == 2:
text += '↑'
if i<len(marks):
text += unidecode(marks[i]).replace(' ','')
return text
def latin_to_hangul(text):
for regex, replacement in _latin_to_hangul:
text = re.sub(regex, replacement, text)
return text
def divide_hangul(text):
for regex, replacement in _hangul_divided:
text = re.sub(regex, replacement, text)
return text
def hangul_number(num, sino=True):
'''Reference https://github.com/Kyubyong/g2pK'''
num = re.sub(',', '', num)
if num == '0':
return '영'
if not sino and num == '20':
return '슀무'
digits = '123456789'
names = 'μΌμ΄μ‚Όμ‚¬μ˜€μœ‘μΉ νŒ”κ΅¬'
digit2name = {d: n for d, n in zip(digits, names)}
modifiers = 'ν•œ 두 μ„Έ λ„€ λ‹€μ„― μ—¬μ„― 일곱 μ—¬λŸ 아홉'
decimals = 'μ—΄ 슀물 μ„œλ₯Έ λ§ˆν” μ‰° 예순 일흔 μ—¬λ“  아흔'
digit2mod = {d: mod for d, mod in zip(digits, modifiers.split())}
digit2dec = {d: dec for d, dec in zip(digits, decimals.split())}
spelledout = []
for i, digit in enumerate(num):
i = len(num) - i - 1
if sino:
if i == 0:
name = digit2name.get(digit, '')
elif i == 1:
name = digit2name.get(digit, '') + 'μ‹­'
name = name.replace('일십', 'μ‹­')
else:
if i == 0:
name = digit2mod.get(digit, '')
elif i == 1:
name = digit2dec.get(digit, '')
if digit == '0':
if i % 4 == 0:
last_three = spelledout[-min(3, len(spelledout)):]
if ''.join(last_three) == '':
spelledout.append('')
continue
else:
spelledout.append('')
continue
if i == 2:
name = digit2name.get(digit, '') + 'λ°±'
name = name.replace('일백', '백')
elif i == 3:
name = digit2name.get(digit, '') + '천'
name = name.replace('일천', '천')
elif i == 4:
name = digit2name.get(digit, '') + '만'
name = name.replace('일만', '만')
elif i == 5:
name = digit2name.get(digit, '') + 'μ‹­'
name = name.replace('일십', 'μ‹­')
elif i == 6:
name = digit2name.get(digit, '') + 'λ°±'
name = name.replace('일백', '백')
elif i == 7:
name = digit2name.get(digit, '') + '천'
name = name.replace('일천', '천')
elif i == 8:
name = digit2name.get(digit, '') + 'μ–΅'
elif i == 9:
name = digit2name.get(digit, '') + 'μ‹­'
elif i == 10:
name = digit2name.get(digit, '') + 'λ°±'
elif i == 11:
name = digit2name.get(digit, '') + '천'
elif i == 12:
name = digit2name.get(digit, '') + 'μ‘°'
elif i == 13:
name = digit2name.get(digit, '') + 'μ‹­'
elif i == 14:
name = digit2name.get(digit, '') + 'λ°±'
elif i == 15:
name = digit2name.get(digit, '') + '천'
spelledout.append(name)
return ''.join(elem for elem in spelledout)
def number_to_hangul(text):
'''Reference https://github.com/Kyubyong/g2pK'''
tokens = set(re.findall(r'(\d[\d,]*)([\uac00-\ud71f]+)', text))
for token in tokens:
num, classifier = token
if classifier[:2] in _korean_classifiers or classifier[0] in _korean_classifiers:
spelledout = hangul_number(num, sino=False)
else:
spelledout = hangul_number(num, sino=True)
text = text.replace(f'{num}{classifier}', f'{spelledout}{classifier}')
# digit by digit for remaining digits
digits = '0123456789'
names = 'μ˜μΌμ΄μ‚Όμ‚¬μ˜€μœ‘μΉ νŒ”κ΅¬'
for d, n in zip(digits, names):
text = text.replace(d, n)
return text
def number_to_chinese(text):
numbers = re.findall(r'\d+(?:\.?\d+)?', text)
for number in numbers:
text = text.replace(number, cn2an.an2cn(number),1)
return text
def chinese_to_bopomofo(text):
text=text.replace('、',',').replace('οΌ›',',').replace(':',',')
words=jieba.lcut(text,cut_all=False)
text=''
for word in words:
bopomofos=lazy_pinyin(word,BOPOMOFO)
if not re.search('[\u4e00-\u9fff]',word):
text+=word
continue
for i in range(len(bopomofos)):
if re.match('[\u3105-\u3129]',bopomofos[i][-1]):
bopomofos[i]+='Λ‰'
if text!='':
text+=' '
text+=''.join(bopomofos)
return text
def latin_to_bopomofo(text):
for regex, replacement in _latin_to_bopomofo:
text = re.sub(regex, replacement, text)
return text
def bopomofo_to_romaji(text):
for regex, replacement in _bopomofo_to_romaji:
text = re.sub(regex, replacement, text)
return text
def basic_cleaners(text):
'''Basic pipeline that lowercases and collapses whitespace without transliteration.'''
text = lowercase(text)
text = collapse_whitespace(text)
return text
def transliteration_cleaners(text):
'''Pipeline for non-English text that transliterates to ASCII.'''
text = convert_to_ascii(text)
text = lowercase(text)
text = collapse_whitespace(text)
return text
def japanese_cleaners(text):
text=japanese_to_romaji_with_accent(text)
if re.match('[A-Za-z]',text[-1]):
text += '.'
return text
def japanese_cleaners2(text):
return japanese_cleaners(text).replace('ts','Κ¦').replace('...','…')
def korean_cleaners(text):
'''Pipeline for Korean text'''
text = latin_to_hangul(text)
text = number_to_hangul(text)
text = j2hcj(h2j(text))
text = divide_hangul(text)
if re.match('[\u3131-\u3163]',text[-1]):
text += '.'
return text
def chinese_cleaners(text):
'''Pipeline for Chinese text'''
text=number_to_chinese(text)
text=chinese_to_bopomofo(text)
text=latin_to_bopomofo(text)
if re.match('[Λ‰ΛŠΛ‡Λ‹Λ™]',text[-1]):
text += '。'
return text
def zh_ja_mixture_cleaners(text):
chinese_texts=re.findall(r'\[ZH\].*?\[ZH\]',text)
japanese_texts=re.findall(r'\[JA\].*?\[JA\]',text)
for chinese_text in chinese_texts:
cleaned_text=number_to_chinese(chinese_text[4:-4])
cleaned_text=chinese_to_bopomofo(cleaned_text)
cleaned_text=latin_to_bopomofo(cleaned_text)
cleaned_text=bopomofo_to_romaji(cleaned_text)
cleaned_text=re.sub('i[aoe]',lambda x:'y'+x.group(0)[1:],cleaned_text)
cleaned_text=re.sub('u[aoΙ™e]',lambda x:'w'+x.group(0)[1:],cleaned_text)
cleaned_text=re.sub('([Κ¦sΙΉ]`[⁼ʰ]?)([→↓↑]+)',lambda x:x.group(1)+'ΙΉ`'+x.group(2),cleaned_text).replace('Ι»','ΙΉ`')
cleaned_text=re.sub('([Κ¦s][⁼ʰ]?)([→↓↑]+)',lambda x:x.group(1)+'ΙΉ'+x.group(2),cleaned_text)
text = text.replace(chinese_text,cleaned_text+' ',1)
for japanese_text in japanese_texts:
cleaned_text=japanese_to_romaji_with_accent(japanese_text[4:-4]).replace('ts','Κ¦').replace('u','Ι―').replace('...','…')
text = text.replace(japanese_text,cleaned_text+' ',1)
text=text[:-1]
if re.match('[A-Za-zΙ―ΙΉΙ™Ι₯→↓↑]',text[-1]):
text += '.'
return text