updated requirements

utils/__init__.py

@@ -1,5 +1,11 @@
 import torch
 import torch.nn as nn
+# from icefall.utils import make_pad_mask
+
+from .symbol_table import SymbolTable
+
+# make_pad_mask = make_pad_mask
+SymbolTable = SymbolTable
 
 
 class Transpose(nn.Identity):

utils/g2p/__init__.py

@@ -14,14 +14,15 @@ class PhonemeBpeTokenizer:
 
   def tokenize(self, text):
     # 1. convert text to phoneme
-    phonemes = _clean_text(text, ['cje_cleaners'])
+    phonemes, langs = _clean_text(text, ['cje_cleaners'])
     # 2. replace blank space " " with "_"
     phonemes = phonemes.replace(" ", "_")
     # 3. tokenize phonemes
     phoneme_tokens = self.tokenizer.encode(phonemes).ids
+    assert(len(phoneme_tokens) == len(langs))
     if not len(phoneme_tokens):
-      phoneme_tokens = self.tokenizer.encode(text).ids
-    return phoneme_tokens
+      raise ValueError("Empty text is given")
+    return phoneme_tokens, langs
 
 def text_to_sequence(text, cleaner_names):
   '''Converts a string of text to a sequence of IDs corresponding to the symbols in the text.
@@ -67,5 +68,5 @@ def _clean_text(text, cleaner_names):
     cleaner = getattr(cleaners, name)
     if not cleaner:
       raise Exception('Unknown cleaner: %s' % name)
-    text = cleaner(text)
-  return text
+    text, langs = cleaner(text)
+  return text, langs

utils/g2p/bpe_1024.json

@@ -0,0 +1,2049 @@
+{
+  "version": "1.0",
+  "truncation": null,
+  "padding": null,
+  "added_tokens": [
+    {
+      "id": 0,
+      "content": "[UNK]",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    {
+      "id": 1,
+      "content": "[CLS]",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    {
+      "id": 2,
+      "content": "[SEP]",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    {
+      "id": 3,
+      "content": "[PAD]",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    },
+    {
+      "id": 4,
+      "content": "[MASK]",
+      "single_word": false,
+      "lstrip": false,
+      "rstrip": false,
+      "normalized": false,
+      "special": true
+    }
+  ],
+  "normalizer": null,
+  "pre_tokenizer": {
+    "type": "Whitespace"
+  },
+  "post_processor": null,
+  "decoder": null,
+  "model": {
+    "type": "BPE",
+    "dropout": null,
+    "unk_token": "[UNK]",
+    "continuing_subword_prefix": null,
+    "end_of_word_suffix": null,
+    "fuse_unk": false,
+    "byte_fallback": false,
+    "vocab": {
+      "[UNK]": 0,
+      "[CLS]": 1,
+      "[SEP]": 2,
+      "[PAD]": 3,
+      "[MASK]": 4,
+      "!": 5,
+      "#": 6,
+      "*": 7,
+      ",": 8,
+      "-": 9,
+      ".": 10,
+      "=": 11,
+      "?": 12,
+      "N": 13,
+      "Q": 14,
+      "^": 15,
+      "_": 16,
+      "`": 17,
+      "a": 18,
+      "b": 19,
+      "d": 20,
+      "e": 21,
+      "f": 22,
+      "g": 23,
+      "h": 24,
+      "i": 25,
+      "j": 26,
+      "k": 27,
+      "l": 28,
+      "m": 29,
+      "n": 30,
+      "o": 31,
+      "p": 32,
+      "s": 33,
+      "t": 34,
+      "u": 35,
+      "v": 36,
+      "w": 37,
+      "x": 38,
+      "y": 39,
+      "z": 40,
+      "~": 41,
+      "æ": 42,
+      "ç": 43,
+      "ð": 44,
+      "ŋ": 45,
+      "ɑ": 46,
+      "ɔ": 47,
+      "ə": 48,
+      "ɛ": 49,
+      "ɥ": 50,
+      "ɪ": 51,
+      "ɫ": 52,
+      "ɯ": 53,
+      "ɸ": 54,
+      "ɹ": 55,
+      "ɾ": 56,
+      "ʃ": 57,
+      "ʊ": 58,
+      "ʑ": 59,
+      "ʒ": 60,
+      "ʰ": 61,
+      "ˈ": 62,
+      "ˌ": 63,
+      "θ": 64,
+      "…": 65,
+      "⁼": 66,
+      "↑": 67,
+      "→": 68,
+      "↓": 69,
+      "_t": 70,
+      "↓↑": 71,
+      "_ˈ": 72,
+      "ən": 73,
+      "_s": 74,
+      "aɪ": 75,
+      "əɹ": 76,
+      "eɪ": 77,
+      "oʊ": 78,
+      "_k": 79,
+      "ʃi": 80,
+      "_w": 81,
+      "_ð": 82,
+      "ts": 83,
+      "tʃ": 84,
+      "_ts": 85,
+      "_h": 86,
+      "_ə": 87,
+      "_m": 88,
+      "an": 89,
+      "_n": 90,
+      "_ðə": 91,
+      "ɛn": 92,
+      "ɑʊ": 93,
+      "ɑŋ": 94,
+      "`⁼": 95,
+      "_p": 96,
+      "_i": 97,
+      "_ɪ": 98,
+      "_tʃ": 99,
+      "_l": 100,
+      "jɛn": 101,
+      "_d": 102,
+      "_f": 103,
+      "_j": 104,
+      "wo": 105,
+      "_b": 106,
+      "ta": 107,
+      "`↓": 108,
+      "te": 109,
+      "ənd": 110,
+      "_ʃi": 111,
+      "wa": 112,
+      "ka": 113,
+      "ɪŋ": 114,
+      "in": 115,
+      "st": 116,
+      "li": 117,
+      "ʊŋ": 118,
+      "_tɪ": 119,
+      "to": 120,
+      "weɪ": 121,
+      "_ənd": 122,
+      "ʰi": 123,
+      "_əv": 124,
+      "əŋ": 125,
+      "no": 126,
+      "_x": 127,
+      "ɾɯ": 128,
+      "na": 129,
+      "_a": 130,
+      "_ɹ": 131,
+      "ɪn": 132,
+      "ga": 133,
+      "de": 134,
+      "joʊ": 135,
+      "æn": 136,
+      "kɯ": 137,
+      "ɾe": 138,
+      "ma": 139,
+      "_ðə_ˈ": 140,
+      "ɾa": 141,
+      "ɛɹ": 142,
+      "mo": 143,
+      "ɔɹ": 144,
+      "əɫ": 145,
+      "_g": 146,
+      "da": 147,
+      "*↑": 148,
+      "ɪˈ": 149,
+      "_o": 150,
+      "_ʃ": 151,
+      "iŋ": 152,
+      "ja": 153,
+      "əm": 154,
+      "_ˌ": 155,
+      "aʊ": 156,
+      "_əˈ": 157,
+      "`↑": 158,
+      "ət": 159,
+      "_aɪ": 160,
+      "oo": 161,
+      "sɯ": 162,
+      "↓.": 163,
+      "_ɪn": 164,
+      "_hi": 165,
+      "_wɪ": 166,
+      "ɪz": 167,
+      "_na": 168,
+      "wan": 169,
+      "_ko": 170,
+      "_wo": 171,
+      "ɪd": 172,
+      "ɾi": 173,
+      "_ju": 174,
+      "mə": 175,
+      "_lə": 176,
+      "_hæ": 177,
+      "_ðət": 178,
+      "ɑɹ": 179,
+      "tʰ": 180,
+      "ki": 181,
+      "……": 182,
+      "ɑz": 183,
+      "_ɔ": 184,
+      "_mi": 185,
+      "_wɑz": 186,
+      "_ˈs": 187,
+      "↓,": 188,
+      "_tʰ": 189,
+      "əˈ": 190,
+      "dʑ": 191,
+      "ɪt": 192,
+      "_kʰ": 193,
+      "iɛ": 194,
+      "_ma": 195,
+      "ɪs": 196,
+      "tsɯ": 197,
+      "_ni": 198,
+      "_ɪt": 199,
+      "ke": 200,
+      "iɑʊ": 201,
+      "_ka": 202,
+      "_əɹ": 203,
+      "nd": 204,
+      "_ˈp": 205,
+      "ko": 206,
+      "jo": 207,
+      "ɹi": 208,
+      "mən": 209,
+      "ʊd": 210,
+      "_ˈm": 211,
+      "_fəɹ": 212,
+      "tʃʰi": 213,
+      "sa": 214,
+      "ʰɥ": 215,
+      "kʰ": 216,
+      "ˈs": 217,
+      "ɑt": 218,
+      "ɛd": 219,
+      "se": 220,
+      "tʃi": 221,
+      "ɛɫ": 222,
+      "_ˈk": 223,
+      "_joʊ": 224,
+      "təɹ": 225,
+      "ɛz": 226,
+      "--": 227,
+      "vəɹ": 228,
+      "`→": 229,
+      "ʃən": 230,
+      "_ɪz": 231,
+      "_meɪ": 232,
+      "_æ": 233,
+      "dʒ": 234,
+      "_ki": 235,
+      "_hɪz": 236,
+      "_bi": 237,
+      "uɑŋ": 238,
+      "_ˈf": 239,
+      "↓↑.": 240,
+      "_wɪθ": 241,
+      "ju": 242,
+      "iɑŋ": 243,
+      "→.": 244,
+      "_so": 245,
+      "_həɹ": 246,
+      "↑.": 247,
+      "ni": 248,
+      "_mo": 249,
+      "_maɪ": 250,
+      "laɪ": 251,
+      "ɥɛ": 252,
+      "_ta": 253,
+      "ənt": 254,
+      "_tʃʰi": 255,
+      "_sɯ": 256,
+      "_θ": 257,
+      "_ɛz": 258,
+      "wən": 259,
+      "me": 260,
+      "mi": 261,
+      "_hæd": 262,
+      "_ha": 263,
+      "əs": 264,
+      "_ˈl": 265,
+      "_st": 266,
+      "ðəɹ": 267,
+      "oʊn": 268,
+      "_wa": 269,
+      "ʰəŋ": 270,
+      "_nɑt": 271,
+      "*.": 272,
+      "kt": 273,
+      "_ˈh": 274,
+      "do": 275,
+      "ɥæn": 276,
+      "ne": 277,
+      "_to": 278,
+      "_wən": 279,
+      "_no": 280,
+      "_laɪ": 281,
+      "_wəɹ": 282,
+      "↑,": 283,
+      "→,": 284,
+      "ɛs": 285,
+      "↓↑,": 286,
+      "_ɔn": 287,
+      "ʰu": 288,
+      "so": 289,
+      "_ˈb": 290,
+      "ɫd": 291,
+      "ɪk": 292,
+      "ɪst": 293,
+      "_fɹ": 294,
+      "_ðɛɹ": 295,
+      "_weɪ": 296,
+      "kaɾa": 297,
+      "_ˈd": 298,
+      "_hæv": 299,
+      "tsʰ": 300,
+      "waɪ": 301,
+      "ɾo": 302,
+      "ɛm": 303,
+      "_æt": 304,
+      "ʊɹ": 305,
+      "_ˈw": 306,
+      "ba": 307,
+      "_noʊ": 308,
+      "ʰjɛn": 309,
+      "ɹeɪ": 310,
+      "_jo": 311,
+      "ɸɯ": 312,
+      "_sa": 313,
+      "_ɹɪˈ": 314,
+      "_ˈn": 315,
+      "ai": 316,
+      "_bət": 317,
+      "ɪɹ": 318,
+      "tʃʰɥ": 319,
+      "_dʑ": 320,
+      "əˌ": 321,
+      "_ðɪs": 322,
+      "..": 323,
+      "xwa": 324,
+      "_ɪm": 325,
+      "_dɪˈ": 326,
+      "_kən": 327,
+      "dʑi": 328,
+      "*,": 329,
+      "ɑn": 330,
+      "_ʃiɑŋ": 331,
+      "_kɯ": 332,
+      "ʃin": 333,
+      "_soʊ": 334,
+      "bi": 335,
+      "tʰjɛn": 336,
+      "te_i": 337,
+      "_tsʰ": 338,
+      "_ɯ": 339,
+      "aɪt": 340,
+      "ʰiŋ": 341,
+      "ðə": 342,
+      "_ɔɫ": 343,
+      "_ˈɹ": 344,
+      "nai": 345,
+      "əɹd": 346,
+      "_ˈt": 347,
+      "_ən": 348,
+      "_tʃʰɥ": 349,
+      "_iɛ": 350,
+      "leɪ": 351,
+      "ɛɹi": 352,
+      "ˈt": 353,
+      "ha": 354,
+      "ʃiŋ": 355,
+      "ɛvəɹ": 356,
+      "zɯ": 357,
+      "_wi": 358,
+      "_ja": 359,
+      "ɛk": 360,
+      "ʰɑŋ": 361,
+      "_tsɯ": 362,
+      "_əv_ðə": 363,
+      "taʃi": 364,
+      "_sɛd": 365,
+      "_xə": 366,
+      "_li": 367,
+      "_si": 368,
+      "desɯ": 369,
+      "_ˌɪn": 370,
+      "ʃjɛn": 371,
+      "_baɪ": 372,
+      "on": 373,
+      "_xɑʊ": 374,
+      "_ðeɪ": 375,
+      "_xaɪ": 376,
+      "`↓↑": 377,
+      "xweɪ": 378,
+      "hi": 379,
+      "_se": 380,
+      "ə_s": 381,
+      "_fɹəm": 382,
+      "ʊt": 383,
+      "di": 384,
+      "aʊt": 385,
+      "əb": 386,
+      "sɹ": 387,
+      "əz": 388,
+      "_xweɪ": 389,
+      "_kʰə": 390,
+      "ɹu": 391,
+      "_u": 392,
+      "_de": 393,
+      "aɪd": 394,
+      "ɪv": 395,
+      "bɯ": 396,
+      "_ho": 397,
+      "əɹz": 398,
+      "joo": 399,
+      "_bɪˈ": 400,
+      "_tʰa": 401,
+      "ɛt": 402,
+      "en": 403,
+      "ɛni": 404,
+      "əst": 405,
+      "æk": 406,
+      "ə_ts": 407,
+      "_ˈɪn": 408,
+      "ti": 409,
+      "ɥn": 410,
+      "_dʒ": 411,
+      "xɑʊ": 412,
+      "_ˈv": 413,
+      "ʃiɑŋ": 414,
+      "pʰ": 415,
+      "_wɪtʃ": 416,
+      "eɪm": 417,
+      "oʊz": 418,
+      "əðəɹ": 419,
+      "fɑŋ": 420,
+      "_ˈg": 421,
+      "_do": 422,
+      "_ʃiɑʊ": 423,
+      "_ˈæ": 424,
+      "_jʊɹ": 425,
+      "_ðɛm": 426,
+      "ɪm": 427,
+      "ɛst": 428,
+      "ænd": 429,
+      "_du": 430,
+      "ɯɯ": 431,
+      "kan": 432,
+      "_da": 433,
+      "ino": 434,
+      "_e": 435,
+      "_wʊd": 436,
+      "ɛnd": 437,
+      "meɪ": 438,
+      "θɪŋ": 439,
+      "_ʃjɛn": 440,
+      "iz": 441,
+      "aɪm": 442,
+      "_hu": 443,
+      "_əˈb": 444,
+      "əns": 445,
+      "_wɪɫ": 446,
+      "tʰi": 447,
+      "go": 448,
+      "ɛnt": 449,
+      "fu": 450,
+      "æp": 451,
+      "xoʊ": 452,
+      "eɪk": 453,
+      "ʊk": 454,
+      "əɹˈ": 455,
+      "_θɪŋ": 456,
+      "əl": 457,
+      "pɹ": 458,
+      "ətʃ": 459,
+      "nt": 460,
+      "_ɸɯ": 461,
+      "lu": 462,
+      "_ˈɔ": 463,
+      "_iɑʊ": 464,
+      "lə": 465,
+      "tu": 466,
+      "_dʑi": 467,
+      "eɪt": 468,
+      "_ʃin": 469,
+      "nna": 470,
+      "_ˈpɹ": 471,
+      "fən": 472,
+      "_əp": 473,
+      "njɛn": 474,
+      "_aʊt": 475,
+      "fɔɹ": 476,
+      "_tu": 477,
+      "eɪʃən": 478,
+      "ɪɫ": 479,
+      "_wət": 480,
+      "_ɪf": 481,
+      "_ɥ": 482,
+      "_fa": 483,
+      "ˈw": 484,
+      "tʃʰjɛn": 485,
+      "_wɪn": 486,
+      "oʊɫd": 487,
+      "_əˈp": 488,
+      "aʊnd": 489,
+      "san": 490,
+      "he": 491,
+      "_bɪn": 492,
+      "fa": 493,
+      "ɪf": 494,
+      "ɔŋ": 495,
+      "ge": 496,
+      "_ɪn_ðə": 497,
+      "miŋ": 498,
+      "_pɹ": 499,
+      "ina": 500,
+      "ano": 501,
+      "əbəɫ": 502,
+      "kˈs": 503,
+      "_ˈɛni": 504,
+      "nəŋ": 505,
+      "əd": 506,
+      "_əv_ðə_ˈ": 507,
+      "_waɪ": 508,
+      "_taɪm": 509,
+      "ˈsɛɫ": 510,
+      "ʃiɛ": 511,
+      "_kəm": 512,
+      "æst": 513,
+      "_goʊ": 514,
+      "mɯ": 515,
+      "ˈp": 516,
+      "_ˈst": 517,
+      "ə_t": 518,
+      "pt": 519,
+      "_pʰ": 520,
+      "ʰɹ": 521,
+      "ʃja": 522,
+      "iwa": 523,
+      "ɪl": 524,
+      "bət": 525,
+      "_fɑŋ": 526,
+      "ho": 527,
+      "iv": 528,
+      "loʊ": 529,
+      "be": 530,
+      "_laɪk": 531,
+      "ɪʃ": 532,
+      "_fu": 533,
+      "ze": 534,
+      "ə_tʃ": 535,
+      "ɑɹt": 536,
+      "ɔɹd": 537,
+      "tʃʰiŋ": 538,
+      "mp": 539,
+      "_ðə_s": 540,
+      "_əˈbaʊt": 541,
+      "_ˈoʊ": 542,
+      "kʰə": 543,
+      "d_tɪ": 544,
+      "ŋga": 545,
+      "əli": 546,
+      "_kʰan": 547,
+      "çi": 548,
+      "_ˈju": 549,
+      "_kʊd": 550,
+      "ɔɫ": 551,
+      "ɔt": 552,
+      "_ɪts": 553,
+      "_san": 554,
+      "tʃa": 555,
+      "i_na": 556,
+      "xə": 557,
+      "ɛkt": 558,
+      "_mɔɹ": 559,
+      "te_kɯ": 560,
+      "ɪdʒ": 561,
+      "jʊŋ": 562,
+      "_wan": 563,
+      "æt": 564,
+      "kat": 565,
+      "ˈsɛɫf": 566,
+      "_ke": 567,
+      "aɪnd": 568,
+      "it": 569,
+      "_ɑɹ": 570,
+      "sp": 571,
+      "oʊnt": 572,
+      "_tʃi": 573,
+      "tsʰɹ": 574,
+      "_xən": 575,
+      "_əˈg": 576,
+      "ə_k": 577,
+      "to_i": 578,
+      "_tʰi": 579,
+      "_iŋ": 580,
+      "aʊn": 581,
+      "gɯ": 582,
+      "_ɪkˈs": 583,
+      "ɛv": 584,
+      "gi": 585,
+      "ks": 586,
+      "_səm": 587,
+      "ana": 588,
+      "ɪtəɫ": 589,
+      "nan": 590,
+      "_ˈɪntu": 591,
+      "_hiɹ": 592,
+      "_te": 593,
+      "_naʊ": 594,
+      "ʃiɑʊ": 595,
+      "ʃo": 596,
+      "ɹe": 597,
+      "xaɪ": 598,
+      "_tʃʰiŋ": 599,
+      "_sɹ": 600,
+      "_haʊ": 601,
+      "?.": 602,
+      "_feɪ": 603,
+      "liŋ": 604,
+      "_ʃja": 605,
+      "_ˈdʒ": 606,
+      "_seɪ": 607,
+      "ˈn": 608,
+      "soʊ": 609,
+      "tʰʊŋ": 610,
+      "_ljoʊ": 611,
+      "maɪ": 612,
+      "_bɹ": 613,
+      "ɹeɪt": 614,
+      "_nəŋ": 615,
+      "ʰə": 616,
+      "æns": 617,
+      "_ˈɔl": 618,
+      "tatʃi": 619,
+      "nto": 620,
+      "_ˌɪnˈ": 621,
+      "le": 622,
+      "nde": 623,
+      "_ˈvɛɹi": 624,
+      "mənt": 625,
+      "ɾima": 626,
+      "_ðɛn": 627,
+      "_həz": 628,
+      "_ɹi": 629,
+      "ftəɹ": 630,
+      "_sp": 631,
+      "ɾewa": 632,
+      "ga_a": 633,
+      "z_əv": 634,
+      "_miŋ": 635,
+      "_tɪ_ðə": 636,
+      "ɹaɪ": 637,
+      "ɛl": 638,
+      "ɹæ": 639,
+      "_hoʊ": 640,
+      "xu": 641,
+      "oʊnli": 642,
+      "ŋk": 643,
+      "i_i": 644,
+      "_dɪd": 645,
+      "_dʒɪst": 646,
+      "ing": 647,
+      "kai": 648,
+      "_mæn": 649,
+      "_in": 650,
+      "zo": 651,
+      "əf": 652,
+      "dake": 653,
+      "_ˈsəm": 654,
+      "ɾɯ_no": 655,
+      "_go": 656,
+      "tʃəɹ": 657,
+      "ite": 658,
+      "`↓.": 659,
+      "_kʰaɪ": 660,
+      "sk": 661,
+      "ɔɹs": 662,
+      "_tʰiŋ": 663,
+      "_nə": 664,
+      "pəɫ": 665,
+      "_tɪ_bi": 666,
+      "ˈfɔɹ": 667,
+      "mu": 668,
+      "su": 669,
+      "aa": 670,
+      "ɪstəɹ": 671,
+      "ʰan": 672,
+      "pəɹ": 673,
+      "ə_p": 674,
+      "liɑŋ": 675,
+      "_v": 676,
+      "oʊst": 677,
+      "_əˈgɛn": 678,
+      "ənz": 679,
+      "No": 680,
+      "ɔɹt": 681,
+      "_səˈ": 682,
+      "_mɯ": 683,
+      "tʃʰ": 684,
+      "_ˈlɪtəɫ": 685,
+      "_xwo": 686,
+      "_ˌbi": 687,
+      "_ˈoʊvəɹ": 688,
+      "_çi": 689,
+      "_deɪ": 690,
+      "aɪn": 691,
+      "_ʃiŋ": 692,
+      "i_ʃi": 693,
+      "_tsʰaɪ": 694,
+      "ʃoo": 695,
+      "ɾoo": 696,
+      "bəɹ": 697,
+      "ʰa": 698,
+      "ˈɛs": 699,
+      "_ɪn_ðə_ˈ": 700,
+      "Nwa": 701,
+      "_ðən": 702,
+      "saɪ": 703,
+      "_ˈjuˈɛs": 704,
+      "nda": 705,
+      "_pleɪ": 706,
+      "ɪŋ_tɪ": 707,
+      "ɪti": 708,
+      "_me": 709,
+      "_ʃʊd": 710,
+      "_nu": 711,
+      "_ðə_k": 712,
+      "za": 713,
+      "_ˈɛvəɹ": 714,
+      "əɹn": 715,
+      "æd": 716,
+      "ˈm": 717,
+      "_doʊnt": 718,
+      "_məst": 719,
+      "jɯɯ": 720,
+      "ɑɹd": 721,
+      "_jɛn": 722,
+      "ʃɥ": 723,
+      "_ˈoʊnli": 724,
+      "_ʃo": 725,
+      "_liŋ": 726,
+      "ss": 727,
+      "ɑl": 728,
+      "dea": 729,
+      "ɾeta": 730,
+      "mjɛn": 731,
+      "_gʊd": 732,
+      "_wɔ": 733,
+      "imo": 734,
+      "no_ko": 735,
+      "_ɥæn": 736,
+      "ndʒ": 737,
+      "ɪʃən": 738,
+      "o_ʃi": 739,
+      "_θɪŋk": 740,
+      "_nan": 741,
+      "to_o": 742,
+      "_tʰʊŋ": 743,
+      "ljoʊ": 744,
+      "tai": 745,
+      "mə_s": 746,
+      "_jɯ": 747,
+      "_uɑŋ": 748,
+      "_ˌbiˈfɔɹ": 749,
+      "æs": 750,
+      "_tʃʰjɛn": 751,
+      "ik": 752,
+      "_bæk": 753,
+      "_ˈiv": 754,
+      "eɪn": 755,
+      "un": 756,
+      "la": 757,
+      "ˈk": 758,
+      "_daʊn": 759,
+      "anai": 760,
+      "_lɛ": 761,
+      "əɹt": 762,
+      "ðɛɹ": 763,
+      "_ˈæftəɹ": 764,
+      "dat": 765,
+      "fan": 766,
+      "bəɫ": 767,
+      "temo": 768,
+      "tʰa": 769,
+      "ɾɯ_ko": 770,
+      "ˈv": 771,
+      "feɪ": 772,
+      "_mətʃ": 773,
+      "xwo": 774,
+      "ɹoʊ": 775,
+      "_ba": 776,
+      "_ˈnɛvəɹ": 777,
+      "_meɪd": 778,
+      "_jʊŋ": 779,
+      "_əˈpɑn": 780,
+      "!?": 781,
+      "_ˈʃ": 782,
+      "_ðə_ˈk": 783,
+      "ft": 784,
+      "_bo": 785,
+      "_ɪn_ə": 786,
+      "tʃʰɥæn": 787,
+      "ˈz": 788,
+      "`↓,": 789,
+      "_bɪˈk": 790,
+      "ɪg": 791,
+      "kin": 792,
+      "_kl": 793,
+      "ɾɯ_n": 794,
+      "_lɑʊ": 795,
+      "----": 796,
+      "ika": 797,
+      "_ɹaɪt": 798,
+      "zd": 799,
+      "z_ənd": 800,
+      "_kjo": 801,
+      "xwan": 802,
+      "too": 803,
+      "_gɪt": 804,
+      "_liɑŋ": 805,
+      "ta_n": 806,
+      "_keɪm": 807,
+      "_ˈəðəɹ": 808,
+      "_wɛɫ": 809,
+      "teki": 810,
+      "see": 811,
+      "jɯ": 812,
+      "i_o": 813,
+      "to_ʃi": 814,
+      "fəɫ": 815,
+      "bo": 816,
+      "ˌt": 817,
+      "ɪp": 818,
+      "ane": 819,
+      "_tʰjɛn": 820,
+      "_tʃo": 821,
+      "ɾjo": 822,
+      "ɪns": 823,
+      "_he": 824,
+      "ŋka": 825,
+      "ʃɥɛ": 826,
+      "dʑa": 827,
+      "vd": 828,
+      "ʰwan": 829,
+      "_gɹeɪt": 830,
+      "_əv_ə": 831,
+      "əndəɹ": 832,
+      "kedo": 833,
+      "_ðə_b": 834,
+      "ək": 835,
+      "_teɪk": 836,
+      "kʰan": 837,
+      "_ˈɔlˌ": 838,
+      "swo": 839,
+      "_ɪt_wɑz": 840,
+      "_ʃɥ": 841,
+      "_sim": 842,
+      "_ˈfɑ": 843,
+      "min": 844,
+      "i_a": 845,
+      "soo": 846,
+      "ɛns": 847,
+      "_sətʃ": 848,
+      "tʰaɪ": 849,
+      "_ga": 850,
+      "i_ka": 851,
+      "koo": 852,
+      "_fəɹst": 853,
+      "_ˈtʃ": 854,
+      "nno": 855,
+      "ə_ɹ": 856,
+      "taɾa": 857,
+      "tʃʰjoʊ": 858,
+      "_æm": 859,
+      "_mu": 860,
+      "_meɪk": 861,
+      "↓…": 862,
+      "ɪˈθ": 863,
+      "ɑb": 864,
+      "ɹa": 865,
+      "_wɛɹ": 866,
+      "_ðə_ˈs": 867,
+      "_əˈl": 868,
+      "_oʊɫd": 869,
+      "æl": 870,
+      "_ˈpi": 871,
+      "_lɔŋ": 872,
+      "dʑo": 873,
+      "_tʰaɪ": 874,
+      "ɔɹn": 875,
+      "əɫz": 876,
+      "_təˈ": 877,
+      "_əˈweɪ": 878,
+      "pa": 879,
+      "_ðiz": 880,
+      "_ˈsp": 881,
+      "nn": 882,
+      "mae": 883,
+      "towa": 884,
+      "ta_no": 885,
+      "_an": 886,
+      "kʰaɪ": 887,
+      "ɾaɾe": 888,
+      "eɪs": 889,
+      "ɑd": 890,
+      "_wɪˈθ": 891,
+      "_ˈivɪn": 892,
+      "_lu": 893,
+      "ɔɪ": 894,
+      "lɪŋ": 895,
+      "əti": 896,
+      "_ðə_f": 897,
+      "oʃi": 898,
+      "_la": 899,
+      "si": 900,
+      "tɪd": 901,
+      "haʊ": 902,
+      "pʰin": 903,
+      "ˈst": 904,
+      "_ˈpəɹ": 905,
+      "eɹ": 906,
+      "*!": 907,
+      "_ˈmɪstəɹ": 908,
+      "ʃa": 909,
+      "_ˌɪm": 910,
+      "ˌθɪŋ": 911,
+      "_neɪ": 912,
+      "_nɥ": 913,
+      "ɑk": 914,
+      "_ɹu": 915,
+      "_ʃɯ": 916,
+      "_ðə_ˈm": 917,
+      "demo": 918,
+      "_dɹ": 919,
+      "dʑoo": 920,
+      "_stɪɫ": 921,
+      "_pʰiŋ": 922,
+      "ə_i": 923,
+      "_ɪkˈsp": 924,
+      "_wɛnt": 925,
+      "ɪɹi": 926,
+      "əˈm": 927,
+      "o_ka": 928,
+      "_əˈk": 929,
+      "ɔk": 930,
+      "_ɥɛ": 931,
+      "_lʊk": 932,
+      "ˈd": 933,
+      "kaʃi": 934,
+      "_wɪθ_ə": 935,
+      "ljɛn": 936,
+      "ɔn": 937,
+      "_ljɛn": 938,
+      "_hɛɫ": 939,
+      "uɹ": 940,
+      "_tʰoʊ": 941,
+      "_tʃʰɥæn": 942,
+      "_sk": 943,
+      "tsʰaɪ": 944,
+      "ɛtəɹ": 945,
+      "_min": 946,
+      "noʊ": 947,
+      "ʃɯ": 948,
+      "_θɹu": 949,
+      "_θɔt": 950,
+      "dajo": 951,
+      "wi": 952,
+      "i_ko": 953,
+      "_tɹ": 954,
+      "_fan": 955,
+      "ɹɛ": 956,
+      "saN": 957,
+      "_hi_wɑz": 958,
+      "_ɾe": 959,
+      "_əm": 960,
+      "te_ki": 961,
+      "_xoʊ": 962,
+      "ˈl": 963,
+      "ˈg": 964,
+      "ga_i": 965,
+      "_ɔn_ðə": 966,
+      "_xwa": 967,
+      "vɪŋ": 968,
+      "man": 969,
+      "fəɹ": 970,
+      "_oʊn": 971,
+      "ˈɹ": 972,
+      "_kɹ": 973,
+      "te_o": 974,
+      "ɪli": 975,
+      "_ʃɥɛ": 976,
+      "_fəŋ": 977,
+      "æɫ": 978,
+      "ɑp": 979,
+      "_ˈɛv": 980,
+      "eɪndʒ": 981,
+      "iɫ": 982,
+      "wət": 983,
+      "ɛðəɹ": 984,
+      "_fən": 985,
+      "ɾee": 986,
+      "_hi_hæd": 987,
+      "_maɪt": 988,
+      "_ge": 989,
+      "ækt": 990,
+      "ɪts": 991,
+      "_hɪm": 992,
+      "_ze": 993,
+      "ii": 994,
+      "_N": 995,
+      "_əv_hɪz": 996,
+      "_gɹ": 997,
+      "ænt": 998,
+      "ɪˌ": 999,
+      "_hɪmˈsɛɫf": 1000,
+      "wa_na": 1001,
+      "aɪəɹ": 1002,
+      "dʑanai": 1003,
+      "kana": 1004,
+      "aɪz": 1005,
+      "_ɪt_ɪz": 1006,
+      "mase": 1007,
+      "wɪn": 1008,
+      "əθɪŋ": 1009,
+      "_pɹəˈ": 1010,
+      "kɯn": 1011,
+      "ˈju": 1012,
+      "_fɔɹ": 1013,
+      "pʰi": 1014,
+      "pʰiŋ": 1015,
+      "o_i": 1016,
+      "vz": 1017,
+      "ɔɪn": 1018,
+      "tʰiŋ": 1019,
+      "_ne": 1020,
+      "gəɹ": 1021,
+      "æts": 1022,
+      "_ˈɹi": 1023
+    },
+    "merges": [
+      "_ t",
+      "↓ ↑",
+      "_ ˈ",
+      "ə n",
+      "_ s",
+      "a ɪ",
+      "ə ɹ",
+      "e ɪ",
+      "o ʊ",
+      "_ k",
+      "ʃ i",
+      "_ w",
+      "_ ð",
+      "t s",
+      "t ʃ",
+      "_t s",
+      "_ h",
+      "_ ə",
+      "_ m",
+      "a n",
+      "_ n",
+      "_ð ə",
+      "ɛ n",
+      "ɑ ʊ",
+      "ɑ ŋ",
+      "` ⁼",
+      "_ p",
+      "_ i",
+      "_ ɪ",
+      "_t ʃ",
+      "_ l",
+      "j ɛn",
+      "_ d",
+      "_ f",
+      "_ j",
+      "w o",
+      "_ b",
+      "t a",
+      "` ↓",
+      "t e",
+      "ən d",
+      "_ ʃi",
+      "w a",
+      "k a",
+      "ɪ ŋ",
+      "i n",
+      "s t",
+      "l i",
+      "ʊ ŋ",
+      "_t ɪ",
+      "t o",
+      "w eɪ",
+      "_ ənd",
+      "ʰ i",
+      "_ə v",
+      "ə ŋ",
+      "n o",
+      "_ x",
+      "ɾ ɯ",
+      "n a",
+      "_ a",
+      "_ ɹ",
+      "ɪ n",
+      "g a",
+      "d e",
+      "j oʊ",
+      "æ n",
+      "k ɯ",
+      "ɾ e",
+      "m a",
+      "_ðə _ˈ",
+      "ɾ a",
+      "ɛ ɹ",
+      "m o",
+      "ɔ ɹ",
+      "ə ɫ",
+      "_ g",
+      "d a",
+      "* ↑",
+      "ɪ ˈ",
+      "_ o",
+      "_ ʃ",
+      "i ŋ",
+      "j a",
+      "ə m",
+      "_ ˌ",
+      "a ʊ",
+      "_ə ˈ",
+      "` ↑",
+      "ə t",
+      "_ aɪ",
+      "o o",
+      "s ɯ",
+      "↓ .",
+      "_ɪ n",
+      "_h i",
+      "_w ɪ",
+      "ɪ z",
+      "_n a",
+      "w an",
+      "_k o",
+      "_w o",
+      "ɪ d",
+      "ɾ i",
+      "_j u",
+      "m ə",
+      "_l ə",
+      "_h æ",
+      "_ðə t",
+      "ɑ ɹ",
+      "t ʰ",
+      "k i",
+      "… …",
+      "ɑ z",
+      "_ ɔ",
+      "_m i",
+      "_w ɑz",
+      "_ˈ s",
+      "↓ ,",
+      "_t ʰ",
+      "ə ˈ",
+      "d ʑ",
+      "ɪ t",
+      "_k ʰ",
+      "i ɛ",
+      "_m a",
+      "ɪ s",
+      "ts ɯ",
+      "_n i",
+      "_ɪ t",
+      "k e",
+      "i ɑʊ",
+      "_k a",
+      "_ əɹ",
+      "n d",
+      "_ˈ p",
+      "k o",
+      "j o",
+      "ɹ i",
+      "m ən",
+      "ʊ d",
+      "_ˈ m",
+      "_f əɹ",
+      "tʃ ʰi",
+      "s a",
+      "ʰ ɥ",
+      "k ʰ",
+      "ˈ s",
+      "ɑ t",
+      "ɛ d",
+      "s e",
+      "t ʃi",
+      "ɛ ɫ",
+      "_ˈ k",
+      "_j oʊ",
+      "t əɹ",
+      "ɛ z",
+      "- -",
+      "v əɹ",
+      "` →",
+      "ʃ ən",
+      "_ɪ z",
+      "_m eɪ",
+      "_ æ",
+      "d ʒ",
+      "_k i",
+      "_h ɪz",
+      "_b i",
+      "u ɑŋ",
+      "_ˈ f",
+      "↓↑ .",
+      "_wɪ θ",
+      "j u",
+      "i ɑŋ",
+      "→ .",
+      "_s o",
+      "_h əɹ",
+      "↑ .",
+      "n i",
+      "_m o",
+      "_m aɪ",
+      "l aɪ",
+      "ɥ ɛ",
+      "_t a",
+      "ən t",
+      "_tʃ ʰi",
+      "_s ɯ",
+      "_ θ",
+      "_ ɛz",
+      "w ən",
+      "m e",
+      "m i",
+      "_hæ d",
+      "_h a",
+      "ə s",
+      "_ˈ l",
+      "_s t",
+      "ð əɹ",
+      "oʊ n",
+      "_w a",
+      "ʰ əŋ",
+      "_n ɑt",
+      "* .",
+      "k t",
+      "_ˈ h",
+      "d o",
+      "ɥ æn",
+      "n e",
+      "_t o",
+      "_w ən",
+      "_n o",
+      "_l aɪ",
+      "_w əɹ",
+      "↑ ,",
+      "→ ,",
+      "ɛ s",
+      "↓↑ ,",
+      "_ɔ n",
+      "ʰ u",
+      "s o",
+      "_ˈ b",
+      "ɫ d",
+      "ɪ k",
+      "ɪ st",
+      "_f ɹ",
+      "_ð ɛɹ",
+      "_w eɪ",
+      "ka ɾa",
+      "_ˈ d",
+      "_hæ v",
+      "ts ʰ",
+      "w aɪ",
+      "ɾ o",
+      "ɛ m",
+      "_æ t",
+      "ʊ ɹ",
+      "_ˈ w",
+      "b a",
+      "_n oʊ",
+      "ʰ jɛn",
+      "ɹ eɪ",
+      "_j o",
+      "ɸ ɯ",
+      "_s a",
+      "_ɹ ɪˈ",
+      "_ˈ n",
+      "a i",
+      "_b ət",
+      "ɪ ɹ",
+      "tʃ ʰɥ",
+      "_d ʑ",
+      "ə ˌ",
+      "_ð ɪs",
+      ". .",
+      "x wa",
+      "_ɪ m",
+      "_d ɪˈ",
+      "_k ən",
+      "dʑ i",
+      "* ,",
+      "ɑ n",
+      "_ʃi ɑŋ",
+      "_k ɯ",
+      "ʃi n",
+      "_s oʊ",
+      "b i",
+      "tʰ jɛn",
+      "te _i",
+      "_ts ʰ",
+      "_ ɯ",
+      "aɪ t",
+      "ʰi ŋ",
+      "ð ə",
+      "_ɔ ɫ",
+      "_ˈ ɹ",
+      "na i",
+      "əɹ d",
+      "_ˈ t",
+      "_ ən",
+      "_tʃ ʰɥ",
+      "_i ɛ",
+      "l eɪ",
+      "ɛɹ i",
+      "ˈ t",
+      "h a",
+      "ʃi ŋ",
+      "ɛ vəɹ",
+      "z ɯ",
+      "_w i",
+      "_j a",
+      "ɛ k",
+      "ʰ ɑŋ",
+      "_ts ɯ",
+      "_əv _ðə",
+      "ta ʃi",
+      "_s ɛd",
+      "_x ə",
+      "_l i",
+      "_s i",
+      "de sɯ",
+      "_ˌ ɪn",
+      "ʃ jɛn",
+      "_b aɪ",
+      "o n",
+      "_x ɑʊ",
+      "_ð eɪ",
+      "_x aɪ",
+      "` ↓↑",
+      "x weɪ",
+      "h i",
+      "_s e",
+      "ə _s",
+      "_fɹ əm",
+      "ʊ t",
+      "d i",
+      "aʊ t",
+      "ə b",
+      "s ɹ",
+      "ə z",
+      "_x weɪ",
+      "_kʰ ə",
+      "ɹ u",
+      "_ u",
+      "_d e",
+      "aɪ d",
+      "ɪ v",
+      "b ɯ",
+      "_h o",
+      "əɹ z",
+      "j oo",
+      "_b ɪˈ",
+      "_tʰ a",
+      "ɛ t",
+      "e n",
+      "ɛn i",
+      "ə st",
+      "æ k",
+      "ə _ts",
+      "_ˈ ɪn",
+      "t i",
+      "ɥ n",
+      "_d ʒ",
+      "x ɑʊ",
+      "_ˈ v",
+      "ʃi ɑŋ",
+      "p ʰ",
+      "_wɪ tʃ",
+      "eɪ m",
+      "oʊ z",
+      "ə ðəɹ",
+      "f ɑŋ",
+      "_ˈ g",
+      "_d o",
+      "_ʃi ɑʊ",
+      "_ˈ æ",
+      "_j ʊɹ",
+      "_ð ɛm",
+      "ɪ m",
+      "ɛ st",
+      "æn d",
+      "_d u",
+      "ɯ ɯ",
+      "k an",
+      "_d a",
+      "in o",
+      "_ e",
+      "_w ʊd",
+      "ɛn d",
+      "m eɪ",
+      "θ ɪŋ",
+      "_ʃ jɛn",
+      "i z",
+      "aɪ m",
+      "_h u",
+      "_əˈ b",
+      "ən s",
+      "_wɪ ɫ",
+      "t ʰi",
+      "g o",
+      "ɛn t",
+      "f u",
+      "æ p",
+      "x oʊ",
+      "eɪ k",
+      "ʊ k",
+      "əɹ ˈ",
+      "_θ ɪŋ",
+      "ə l",
+      "p ɹ",
+      "ə tʃ",
+      "n t",
+      "_ ɸɯ",
+      "l u",
+      "_ˈ ɔ",
+      "_i ɑʊ",
+      "l ə",
+      "t u",
+      "_dʑ i",
+      "eɪ t",
+      "_ʃi n",
+      "n na",
+      "_ˈp ɹ",
+      "f ən",
+      "_ə p",
+      "n jɛn",
+      "_a ʊt",
+      "f ɔɹ",
+      "_t u",
+      "eɪ ʃən",
+      "ɪ ɫ",
+      "_w ət",
+      "_ɪ f",
+      "_ ɥ",
+      "_f a",
+      "ˈ w",
+      "tʃ ʰjɛn",
+      "_w ɪn",
+      "oʊ ɫd",
+      "_əˈ p",
+      "aʊ nd",
+      "s an",
+      "h e",
+      "_b ɪn",
+      "f a",
+      "ɪ f",
+      "ɔ ŋ",
+      "g e",
+      "_ɪn _ðə",
+      "m iŋ",
+      "_p ɹ",
+      "in a",
+      "an o",
+      "əb əɫ",
+      "k ˈs",
+      "_ˈ ɛni",
+      "n əŋ",
+      "ə d",
+      "_əv _ðə_ˈ",
+      "_w aɪ",
+      "_t aɪm",
+      "ˈs ɛɫ",
+      "ʃi ɛ",
+      "_k əm",
+      "æ st",
+      "_g oʊ",
+      "m ɯ",
+      "ˈ p",
+      "_ˈ st",
+      "ə _t",
+      "p t",
+      "_p ʰ",
+      "ʰ ɹ",
+      "ʃ ja",
+      "i wa",
+      "ɪ l",
+      "b ət",
+      "_f ɑŋ",
+      "h o",
+      "i v",
+      "l oʊ",
+      "b e",
+      "_laɪ k",
+      "ɪ ʃ",
+      "_f u",
+      "z e",
+      "ə _tʃ",
+      "ɑɹ t",
+      "ɔɹ d",
+      "tʃʰi ŋ",
+      "m p",
+      "_ðə _s",
+      "_əˈb aʊt",
+      "_ˈ oʊ",
+      "kʰ ə",
+      "d _tɪ",
+      "ŋ ga",
+      "ə li",
+      "_kʰ an",
+      "ç i",
+      "_ˈ ju",
+      "_k ʊd",
+      "ɔ ɫ",
+      "ɔ t",
+      "_ɪ ts",
+      "_s an",
+      "tʃ a",
+      "i _na",
+      "x ə",
+      "ɛ kt",
+      "_m ɔɹ",
+      "te _kɯ",
+      "ɪd ʒ",
+      "j ʊŋ",
+      "_w an",
+      "æ t",
+      "ka t",
+      "ˈsɛɫ f",
+      "_k e",
+      "aɪ nd",
+      "i t",
+      "_ ɑɹ",
+      "s p",
+      "oʊn t",
+      "_t ʃi",
+      "tsʰ ɹ",
+      "_x ən",
+      "_əˈ g",
+      "ə _k",
+      "to _i",
+      "_t ʰi",
+      "_i ŋ",
+      "aʊ n",
+      "g ɯ",
+      "_ɪ kˈs",
+      "ɛ v",
+      "g i",
+      "k s",
+      "_s əm",
+      "an a",
+      "ɪt əɫ",
+      "n an",
+      "_ˈɪn tu",
+      "_hi ɹ",
+      "_t e",
+      "_n aʊ",
+      "ʃi ɑʊ",
+      "ʃ o",
+      "ɹ e",
+      "x aɪ",
+      "_tʃʰi ŋ",
+      "_s ɹ",
+      "_h aʊ",
+      "? .",
+      "_f eɪ",
+      "li ŋ",
+      "_ʃ ja",
+      "_ˈ dʒ",
+      "_s eɪ",
+      "ˈ n",
+      "s oʊ",
+      "tʰ ʊŋ",
+      "_l joʊ",
+      "m aɪ",
+      "_b ɹ",
+      "ɹeɪ t",
+      "_n əŋ",
+      "ʰ ə",
+      "æn s",
+      "_ˈɔ l",
+      "ta tʃi",
+      "n to",
+      "_ˌɪn ˈ",
+      "l e",
+      "n de",
+      "_ˈv ɛɹi",
+      "mən t",
+      "ɾi ma",
+      "_ð ɛn",
+      "_h əz",
+      "_ɹ i",
+      "f təɹ",
+      "_s p",
+      "ɾe wa",
+      "ga _a",
+      "z _əv",
+      "_m iŋ",
+      "_tɪ _ðə",
+      "ɹ aɪ",
+      "ɛ l",
+      "ɹ æ",
+      "_h oʊ",
+      "x u",
+      "oʊn li",
+      "ŋ k",
+      "i _i",
+      "_d ɪd",
+      "_dʒ ɪst",
+      "in g",
+      "ka i",
+      "_m æn",
+      "_i n",
+      "z o",
+      "ə f",
+      "da ke",
+      "_ˈs əm",
+      "ɾɯ _no",
+      "_g o",
+      "tʃ əɹ",
+      "i te",
+      "`↓ .",
+      "_kʰ aɪ",
+      "s k",
+      "ɔɹ s",
+      "_t ʰiŋ",
+      "_n ə",
+      "p əɫ",
+      "_tɪ _bi",
+      "ˈ fɔɹ",
+      "m u",
+      "s u",
+      "a a",
+      "ɪst əɹ",
+      "ʰ an",
+      "p əɹ",
+      "ə _p",
+      "li ɑŋ",
+      "_ v",
+      "oʊ st",
+      "_əˈg ɛn",
+      "ən z",
+      "N o",
+      "ɔɹ t",
+      "_s əˈ",
+      "_m ɯ",
+      "tʃ ʰ",
+      "_ˈl ɪtəɫ",
+      "_x wo",
+      "_ˌ bi",
+      "_ˈoʊ vəɹ",
+      "_ çi",
+      "_d eɪ",
+      "aɪ n",
+      "_ʃi ŋ",
+      "i _ʃi",
+      "_tsʰ aɪ",
+      "ʃ oo",
+      "ɾ oo",
+      "b əɹ",
+      "ʰ a",
+      "ˈ ɛs",
+      "_ɪn _ðə_ˈ",
+      "N wa",
+      "_ð ən",
+      "s aɪ",
+      "_ˈju ˈɛs",
+      "n da",
+      "_p leɪ",
+      "ɪŋ _tɪ",
+      "ɪt i",
+      "_m e",
+      "_ʃ ʊd",
+      "_n u",
+      "_ðə _k",
+      "z a",
+      "_ˈ ɛvəɹ",
+      "əɹ n",
+      "æ d",
+      "ˈ m",
+      "_d oʊnt",
+      "_m əst",
+      "j ɯɯ",
+      "ɑɹ d",
+      "_ jɛn",
+      "ʃ ɥ",
+      "_ˈ oʊnli",
+      "_ʃ o",
+      "_l iŋ",
+      "s s",
+      "ɑ l",
+      "de a",
+      "ɾe ta",
+      "m jɛn",
+      "_g ʊd",
+      "_w ɔ",
+      "i mo",
+      "no _ko",
+      "_ ɥæn",
+      "nd ʒ",
+      "ɪ ʃən",
+      "o _ʃi",
+      "_θɪŋ k",
+      "_n an",
+      "to _o",
+      "_tʰ ʊŋ",
+      "l joʊ",
+      "ta i",
+      "mə _s",
+      "_j ɯ",
+      "_ uɑŋ",
+      "_ˌbi ˈfɔɹ",
+      "æ s",
+      "_tʃ ʰjɛn",
+      "i k",
+      "_b æk",
+      "_ˈ iv",
+      "eɪ n",
+      "u n",
+      "l a",
+      "ˈ k",
+      "_d aʊn",
+      "an ai",
+      "_l ɛ",
+      "əɹ t",
+      "ð ɛɹ",
+      "_ˈæ ftəɹ",
+      "da t",
+      "f an",
+      "b əɫ",
+      "te mo",
+      "tʰ a",
+      "ɾɯ _ko",
+      "ˈ v",
+      "f eɪ",
+      "_m ətʃ",
+      "x wo",
+      "ɹ oʊ",
+      "_b a",
+      "_ˈn ɛvəɹ",
+      "_meɪ d",
+      "_j ʊŋ",
+      "_əˈp ɑn",
+      "! ?",
+      "_ˈ ʃ",
+      "_ðə_ˈ k",
+      "f t",
+      "_b o",
+      "_ɪn _ə",
+      "tʃʰɥ æn",
+      "ˈ z",
+      "`↓ ,",
+      "_bɪˈ k",
+      "ɪ g",
+      "k in",
+      "_k l",
+      "ɾɯ _n",
+      "_l ɑʊ",
+      "-- --",
+      "i ka",
+      "_ɹ aɪt",
+      "z d",
+      "z _ənd",
+      "_k jo",
+      "x wan",
+      "to o",
+      "_g ɪt",
+      "_l iɑŋ",
+      "ta _n",
+      "_k eɪm",
+      "_ˈ əðəɹ",
+      "_w ɛɫ",
+      "te ki",
+      "se e",
+      "j ɯ",
+      "i _o",
+      "to _ʃi",
+      "f əɫ",
+      "b o",
+      "ˌ t",
+      "ɪ p",
+      "an e",
+      "_tʰ jɛn",
+      "_tʃ o",
+      "ɾ jo",
+      "ɪn s",
+      "_h e",
+      "ŋ ka",
+      "ʃ ɥɛ",
+      "dʑ a",
+      "v d",
+      "ʰ wan",
+      "_g ɹeɪt",
+      "_əv _ə",
+      "ənd əɹ",
+      "ke do",
+      "_ðə _b",
+      "ə k",
+      "_t eɪk",
+      "kʰ an",
+      "_ˈɔl ˌ",
+      "s wo",
+      "_ɪt _wɑz",
+      "_ʃ ɥ",
+      "_si m",
+      "_ˈf ɑ",
+      "m in",
+      "i _a",
+      "s oo",
+      "ɛn s",
+      "_s ətʃ",
+      "tʰ aɪ",
+      "_ ga",
+      "i _ka",
+      "k oo",
+      "_fəɹ st",
+      "_ˈ tʃ",
+      "n no",
+      "ə _ɹ",
+      "ta ɾa",
+      "tʃʰ joʊ",
+      "_æ m",
+      "_m u",
+      "_meɪ k",
+      "↓ …",
+      "ɪˈ θ",
+      "ɑ b",
+      "ɹ a",
+      "_w ɛɹ",
+      "_ðə_ˈ s",
+      "_əˈ l",
+      "_ oʊɫd",
+      "æ l",
+      "_ˈp i",
+      "_l ɔŋ",
+      "dʑ o",
+      "_tʰ aɪ",
+      "ɔɹ n",
+      "əɫ z",
+      "_t əˈ",
+      "_əˈ weɪ",
+      "p a",
+      "_ð iz",
+      "_ˈs p",
+      "n n",
+      "ma e",
+      "to wa",
+      "ta _no",
+      "_ an",
+      "kʰ aɪ",
+      "ɾa ɾe",
+      "eɪ s",
+      "ɑ d",
+      "_w ɪˈθ",
+      "_ˈiv ɪn",
+      "_l u",
+      "ɔ ɪ",
+      "l ɪŋ",
+      "ət i",
+      "_ðə _f",
+      "o ʃi",
+      "_l a",
+      "s i",
+      "t ɪd",
+      "h aʊ",
+      "pʰ in",
+      "ˈ st",
+      "_ˈp əɹ",
+      "e ɹ",
+      "* !",
+      "_ˈm ɪstəɹ",
+      "ʃ a",
+      "_ˌ ɪm",
+      "ˌ θɪŋ",
+      "_n eɪ",
+      "_n ɥ",
+      "ɑ k",
+      "_ɹ u",
+      "_ʃ ɯ",
+      "_ðə_ˈ m",
+      "de mo",
+      "_d ɹ",
+      "dʑ oo",
+      "_st ɪɫ",
+      "_p ʰiŋ",
+      "ə _i",
+      "_ɪkˈs p",
+      "_w ɛnt",
+      "ɪ ɹi",
+      "əˈ m",
+      "o _ka",
+      "_əˈ k",
+      "ɔ k",
+      "_ ɥɛ",
+      "_l ʊk",
+      "ˈ d",
+      "ka ʃi",
+      "_wɪθ _ə",
+      "l jɛn",
+      "ɔ n",
+      "_l jɛn",
+      "_h ɛɫ",
+      "u ɹ",
+      "_tʰ oʊ",
+      "_tʃʰɥ æn",
+      "_s k",
+      "tsʰ aɪ",
+      "ɛ təɹ",
+      "_m in",
+      "n oʊ",
+      "ʃ ɯ",
+      "_θ ɹu",
+      "_θ ɔt",
+      "da jo",
+      "w i",
+      "i _ko",
+      "_t ɹ",
+      "_f an",
+      "ɹ ɛ",
+      "sa N",
+      "_hi _wɑz",
+      "_ ɾe",
+      "_ə m",
+      "te _ki",
+      "_x oʊ",
+      "ˈ l",
+      "ˈ g",
+      "ga _i",
+      "_ɔn _ðə",
+      "_x wa",
+      "v ɪŋ",
+      "m an",
+      "f əɹ",
+      "_ oʊn",
+      "ˈ ɹ",
+      "_k ɹ",
+      "te _o",
+      "ɪ li",
+      "_ʃ ɥɛ",
+      "_f əŋ",
+      "æ ɫ",
+      "ɑ p",
+      "_ˈ ɛv",
+      "eɪ ndʒ",
+      "i ɫ",
+      "w ət",
+      "ɛ ðəɹ",
+      "_f ən",
+      "ɾe e",
+      "_hi _hæd",
+      "_maɪ t",
+      "_g e",
+      "æ kt",
+      "ɪ ts",
+      "_h ɪm",
+      "_ ze",
+      "i i",
+      "_ N",
+      "_əv _hɪz",
+      "_g ɹ",
+      "æn t",
+      "ɪ ˌ",
+      "_hɪm ˈsɛɫf",
+      "wa _na",
+      "aɪ əɹ",
+      "dʑ anai",
+      "kan a",
+      "aɪ z",
+      "_ɪt _ɪz",
+      "ma se",
+      "w ɪn",
+      "ə θɪŋ",
+      "_pɹ əˈ",
+      "kɯ n",
+      "ˈ ju",
+      "_f ɔɹ",
+      "p ʰi",
+      "p ʰiŋ",
+      "o _i",
+      "v z",
+      "ɔ ɪn",
+      "t ʰiŋ",
+      "_n e",
+      "g əɹ",
+      "æ ts",
+      "_ˈ ɹi"
+    ]
+  }
+}

utils/g2p/cleaners.py

@@ -2,7 +2,7 @@ import re
 from utils.g2p.japanese import japanese_to_romaji_with_accent, japanese_to_ipa, japanese_to_ipa2, japanese_to_ipa3
 from utils.g2p.mandarin import number_to_chinese, chinese_to_bopomofo, latin_to_bopomofo, chinese_to_romaji, chinese_to_lazy_ipa, chinese_to_ipa, chinese_to_ipa2
 from utils.g2p.english import english_to_lazy_ipa, english_to_ipa2, english_to_lazy_ipa2
-
+patterns = [r'\[EN\](.*?)\[EN\]', r'\[ZH\](.*?)\[ZH\]', r'\[JA\](.*?)\[JA\]']
 def japanese_cleaners(text):
     text = japanese_to_romaji_with_accent(text)
     text = re.sub(r'([A-Za-z])$', r'\1.', text)
@@ -20,6 +20,33 @@ def chinese_cleaners(text):
     return text
 
 def cje_cleaners(text):
+    matches = []
+    for pattern in patterns:
+        matches.extend(re.finditer(pattern, text))
+
+    matches.sort(key=lambda x: x.start())  # Sort matches by their start positions
+
+    outputs = ""
+    output_langs = []
+
+    for match in matches:
+        text_segment = text[match.start():match.end()]
+        phon = clean_one(text_segment)
+        if "[EN]" in text_segment:
+            lang = 'en'
+        elif "[ZH]" in text_segment:
+            lang = 'zh'
+        elif "[JA]" in text_segment:
+            lang = 'ja'
+        else:
+            raise ValueError("If you see this error, please report this bug to issues.")
+        outputs += phon
+        output_langs += [lang] * len(phon)
+    assert len(outputs) == len(output_langs)
+    return outputs, output_langs
+
+
+def clean_one(text):
     if text.find('[ZH]') != -1:
         text = re.sub(r'\[ZH\](.*?)\[ZH\]',
                       lambda x: chinese_to_ipa(x.group(1))+' ', text)