File size: 1,690 Bytes
f643c3e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from text import chinese, japanese, cleaned_text_to_sequence, symbols, english

language_module_map = {"zh": chinese, "ja": japanese, "en": english}
special = [
    ("%", "zh", "SP"),
    ("¥", "zh", "SP2"),
    ("^", "zh", "SP3"),
    # ('@', 'zh', "SP4")#不搞鬼畜了,和第二版保持一致吧
]


def clean_text(text, language):
    for special_s, special_l, target_symbol in special:
        if special_s in text and language == special_l:
            return clean_special(text, language, special_s, target_symbol)
    language_module = language_module_map[language]
    norm_text = language_module.text_normalize(text)
    if language == "zh":
        phones, word2ph = language_module.g2p(norm_text)
        assert len(phones) == sum(word2ph)
        assert len(norm_text) == len(word2ph)
    else:
        phones = language_module.g2p(norm_text)
        word2ph = None

    for ph in phones:
        assert ph in symbols
    return phones, word2ph, norm_text


def clean_special(text, language, special_s, target_symbol):
    """
    特殊静音段sp符号处理
    """
    text = text.replace(special_s, ",")
    language_module = language_module_map[language]
    norm_text = language_module.text_normalize(text)
    phones = language_module.g2p(norm_text)
    new_ph = []
    for ph in phones:
        assert ph in symbols
        if ph == ",":
            new_ph.append(target_symbol)
        else:
            new_ph.append(ph)
    return new_ph


def text_to_sequence(text, language):
    phones = clean_text(text)
    return cleaned_text_to_sequence(phones)


if __name__ == "__main__":
    print(clean_text("你好%啊啊啊额、还是到付红四方。", "zh"))