espnet/text_processing.py at master · cisetn/espnet · GitHub

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
#!/usr/bin/env python3

import re

import jaconv
import numpy as np
import pyopenjtalk

#各条件を正規表現で表す
c1 = '[ウクスツヌフムユルグズヅブプヴ][ァィェォ]'  #ウ段＋「ァ/ィ/ェ/ォ」
c2 = '[イキシチニヒミリギジヂビピ][ャュェョ]'  #イ段（「イ」を除く）＋「ャ/ュ/ェ/ョ」
c3 = '[テデ][ィュ]'  #「テ/デ」＋「ャ/ィ/ュ/ョ」
c4 = '[ァ-ヴー^＾,.、。？！!?]'  #カタカナ１文字（長音含む）

cond = '(' + c1 + '|' + c2 + '|' + c3 + '|' + c4 + ')'
re_mora = re.compile(cond)


def moraWakachi(kana_text):
    return re_mora.findall(kana_text)


def accent_phrase(text):
    import re

    import pyopenjtalk

    phones = []
    prev = -1
    phrase_len = []
    accent_idx = []
    kata = pyopenjtalk.g2p(text, kana=True)

    mora = moraWakachi(kata)
    for labels in pyopenjtalk.run_frontend(text)[1]:
        if labels.split("-")[1].split("+")[0] == "pau":
            phrase_len.append("null")
            accent_idx.append("null")
            continue
        p = re.findall(r"F:([0-9]+)_([0-9]+).*?@([0-9]+)_", labels)
        if len(p) == 1:
            if prev != p[0][2]:
                phrase_len.append(int(p[0][0]))
                accent_idx.append(int(p[0][1]))
            prev = p[0][2]
    left = 0
    phrase = []
    for idx in phrase_len:
        if idx != "null":
            phrase += [mora[left:left + idx]]
            left += idx
        else:
            phrase += [''.join(mora[left:left + 1])]
            left += 1

    return phrase, phrase_len, accent_idx


def text2yomi(text):
    accented = ""
    ap = accent_phrase(text)
    for i, phrase_mora in enumerate(ap[0]):
        if ap[1][i] == "null":
            accented += ''.join(phrase_mora)
            accented += " "
        else:
            phrase = phrase_mora[:ap[2][i]]
            phrase += '＾'
            phrase += phrase_mora[ap[2][i]:]

            accented += ''.join(phrase)
            accented += " "

    yomi = jaconv.kata2hira(accented[0:-1])

    return yomi


def a2p(accented):
    phonemes = np.empty(0)
    print("アクセント句分割：")
    print(accented.split('＃'))
    print("\n")
    for phrase in accented.split('＃'):

        mora = moraWakachi(jaconv.hira2kata(phrase))

        if '^' in mora:
            nucleus = mora.index('^')
            mora.remove('^')
            plane_text = phrase.replace('^', '')
        elif '＾' in mora:
            nucleus = mora.index('＾')
            mora.remove('＾')
            plane_text = phrase.replace('＾', '')
        else:
            phonemes = np.append(phonemes, 'pau')
            continue

        n_mora = len(mora)

        distances = np.arange(1, n_mora + 1) - nucleus
        distances = distances.reshape(-1, 1)

        nucleuses = np.array(nucleus)
        nucleuses = np.tile(nucleuses, (n_mora, 1))

        nuc_dist = np.hstack([nucleuses, distances])
        #print(nuc_dist)

        phrase_phonemes = np.empty(0)
        mora_idx = np.empty(0)
        prev_idx = 0
        i = 0
        print("モーラ分割：")
        print(mora)
        print("\n")

        print("音素 , アクセント核の位置(モーラ単位) , アクセント核からの距離(モーラ単位)：")
        for labels in pyopenjtalk.run_frontend(plane_text)[1]:
            if labels.split("-")[1].split("+")[0] == "pau":
                continue
            p = re.findall(r"\-(.*?)\+.*?\/A:[0-9\-]+\+([0-9\-]+)", labels)
            if len(p) == 1:
                print(p[0][0], end=' , ')
                phrase_phonemes = np.append(phrase_phonemes, p[0][0])
                idx = int(p[0][1]) - 1
                if idx != prev_idx:
                    i += 1
                    prev_idx = idx
                print(nuc_dist[i])

                phrase_phonemes = np.append(phrase_phonemes, nuc_dist[i])

        print("\n")
        phonemes = np.append(phonemes, phrase_phonemes)
        #print(phonemes)
    return phonemes