File size: 8,076 Bytes
7ee3434
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
# Copyright (c) 2023 Amphion.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.


import re
from g2p_en import G2p
from string import punctuation
from typing import Any, Dict, List, Optional, Pattern, Union

from phonemizer.backend import EspeakBackend
from phonemizer.backend.espeak.language_switch import LanguageSwitch
from phonemizer.backend.espeak.words_mismatch import WordMismatch
from phonemizer.punctuation import Punctuation
from phonemizer.separator import Separator

try:
    from pypinyin import Style, pinyin
    from pypinyin.style._utils import get_finals, get_initials
except Exception:
    pass


# This code is modified from
# https://github.com/lifeiteng/vall-e/blob/9c69096d603ce13174fb5cb025f185e2e9b36ac7/valle/data/tokenizer.py


class PypinyinBackend:
    """PypinyinBackend for Chinese. Most codes is referenced from espnet.
    There are two types pinyin or initials_finals, one is
    just like "ni1 hao3", the other is like "n i1 h ao3".
    """

    def __init__(
        self,
        backend="initials_finals",
        punctuation_marks: Union[str, Pattern] = Punctuation.default_marks(),
    ) -> None:
        self.backend = backend
        self.punctuation_marks = punctuation_marks

    def phonemize(
        self, text: List[str], separator: Separator, strip=True, njobs=1
    ) -> List[str]:
        assert isinstance(text, List)
        phonemized = []
        for _text in text:
            _text = re.sub(" +", " ", _text.strip())
            _text = _text.replace(" ", separator.word)
            phones = []
            if self.backend == "pypinyin":
                for n, py in enumerate(
                    pinyin(_text, style=Style.TONE3, neutral_tone_with_five=True)
                ):
                    if all([c in self.punctuation_marks for c in py[0]]):
                        if len(phones):
                            assert phones[-1] == separator.syllable
                            phones.pop(-1)

                        phones.extend(list(py[0]))
                    else:
                        phones.extend([py[0], separator.syllable])
            elif self.backend == "pypinyin_initials_finals":
                for n, py in enumerate(
                    pinyin(_text, style=Style.TONE3, neutral_tone_with_five=True)
                ):
                    if all([c in self.punctuation_marks for c in py[0]]):
                        if len(phones):
                            assert phones[-1] == separator.syllable
                            phones.pop(-1)
                        phones.extend(list(py[0]))
                    else:
                        if py[0][-1].isalnum():
                            initial = get_initials(py[0], strict=False)
                            if py[0][-1].isdigit():
                                final = get_finals(py[0][:-1], strict=False) + py[0][-1]
                            else:
                                final = get_finals(py[0], strict=False)
                            phones.extend(
                                [
                                    initial,
                                    separator.phone,
                                    final,
                                    separator.syllable,
                                ]
                            )
                        else:
                            assert ValueError
            else:
                raise NotImplementedError
            phonemized.append(
                "".join(phones).rstrip(f"{separator.word}{separator.syllable}")
            )
        return phonemized


class G2PModule:
    """Phonemize Text."""

    # We support espeak to extract IPA (International Phonetic Alphabet), which supports 100 languages,
    # https://github.com/espeak-ng/espeak-ng/blob/master/docs/languages.md

    def __init__(
        self,
        language="en-us",
        backend="espeak",
        separator=Separator(word="_", syllable="-", phone="|"),
        preserve_punctuation=True,
        punctuation_marks: Union[str, Pattern] = Punctuation.default_marks(),
        with_stress: bool = False,
        tie: Union[bool, str] = False,
        language_switch: LanguageSwitch = "keep-flags",
        words_mismatch: WordMismatch = "ignore",
    ) -> None:
        self.separator = separator
        self.backend = self._initialize_backend(
            backend,
            language,
            punctuation_marks,
            preserve_punctuation,
            with_stress,
            tie,
            language_switch,
            words_mismatch,
        )

    def _initialize_backend(
        self,
        backend,
        language,
        punctuation_marks,
        preserve_punctuation,
        with_stress,
        tie,
        language_switch,
        words_mismatch,
    ):
        if backend == "espeak":
            return EspeakBackend(
                language,
                punctuation_marks=punctuation_marks,
                preserve_punctuation=preserve_punctuation,
                with_stress=with_stress,
                tie=tie,
                language_switch=language_switch,
                words_mismatch=words_mismatch,
            )
        elif backend in ["pypinyin", "pypinyin_initials_finals"]:
            if language != "cmn":
                raise ValueError(
                    f"{language} is not supported for pypinyin and pypinyin_initials_finals."
                )
            return PypinyinBackend(
                backend=backend,
                punctuation_marks=punctuation_marks + self.separator.word,
            )
        else:
            raise NotImplementedError(f"{backend}")

    def to_list(self, phonemized: str) -> List[str]:
        fields = []
        for word in phonemized.split(self.separator.word):
            pp = re.findall(r"\w+|[^\w\s]", word, re.UNICODE)
            fields.extend(
                [p for p in pp if p != self.separator.phone] + [self.separator.word]
            )
        assert len("".join(fields[:-1])) == len(phonemized) - phonemized.count(
            self.separator.phone
        )
        return fields[:-1]

    def phonemization(self, text, strip=True) -> List[List[str]]:
        if isinstance(text, str):
            text = [text]

        phonemized = self.backend.phonemize(
            text, separator=self.separator, strip=strip, njobs=1
        )
        phonemes = [self.to_list(p) for p in phonemized]
        return phonemes

    def g2p_conversion(self, text: str) -> List[str]:
        phonemes = self.phonemization([text.strip()])
        return phonemes[0]


class LexiconModule:
    def __init__(self, lex_path, language="en-us") -> None:
        # todo: check lexicon derivation, merge with G2PModule?
        lexicon = {}
        with open(lex_path) as f:
            for line in f:
                temp = re.split(r"\s+", line.strip("\n"))
                word = temp[0]
                phones = temp[1:]
                if word.lower() not in lexicon:
                    lexicon[word.lower()] = phones
        self.lexicon = lexicon
        self.language = language
        self.lang2g2p = {"en-us": G2p()}

    def g2p_conversion(self, text):
        phone = None

        # todo: preprocess with other languages
        if self.language == "en-us":
            phone = self.preprocess_english(text)
        else:
            print("No support to", self.language)
            raise

        return phone

    def preprocess_english(self, text):
        text = text.rstrip(punctuation)

        g2p = self.lang2g2p["en-us"]
        phones = []
        words = re.split(r"([,;.\-\?\!\s+])", text)
        for w in words:
            if w.lower() in self.lexicon:
                phones += self.lexicon[w.lower()]
            else:
                phones += list(filter(lambda p: p != " ", g2p(w)))
        phones = "}{".join(phones)
        phones = re.sub(r"\{[^\w\s]?\}", "{sp}", phones)
        phones = phones.replace("}{", " ")

        return phones