phonem updates

.compute

@@ -2,4 +2,4 @@
 source ../tmp/venv/bin/activate
 # python extract_features.py --data_path ${DATA_ROOT}/shared/data/keithito/LJSpeech-1.1/ --cache_path ~/tts_cache/ --config config.json --num_proc 12 --dataset ljspeech --meta_file metadata.csv --val_split 1000 --process_audio true
 # python train.py --config_path config.json --data_path ~/tts_cache/ --debug true
-python train.py --config_path config.json  --data_path ${DATA_ROOT}/shared/data/Blizzard/Nancy/ --debug true
+python train.py --config_path config.json  --data_path ${DATA_ROOT}/shared/data/Blizzard/Nancy/ --debug true

datasets/TTSDataset.py

@@ -6,7 +6,7 @@ import torch
 import random
 from torch.utils.data import Dataset
 
-from utils.text import text_to_sequence, phoneme_to_sequence
+from utils.text import text_to_sequence, phonem_to_sequence
 from utils.data import (prepare_data, pad_per_step, prepare_tensor,
                         prepare_stop_target)
 

utils/text/__init__.py

@@ -3,6 +3,7 @@
 import re
 from utils.text import cleaners
 from utils.text.symbols import symbols, phonemes
+from utils.text.cmudict import text2phone
 
 # Mappings from symbol to numeric ID and vice versa:
 _symbol_to_id = {s: i for i, s in enumerate(symbols)}
@@ -15,9 +16,18 @@ _id_to_phonemes = {i: s for i, s in enumerate(phonemes)}
 _curly_re = re.compile(r'(.*?)\{(.+?)\}(.*)')
 
 
-def phoneme_to_sequence(text, cleaner_names):
+def phonem_to_sequence(text, cleaner_names):
+    '''
+    TODO: This ignores punctuations
+    '''
     sequence = []
-    sequence += _phonem_to_sequence(_clean_text(text, cleaner_names))
+    clean_text = _clean_text(text, cleaner_names)
+    for word in clean_text.split():
+        phonems_text = text2phone(word)
+        if phonems_text == None:
+            continue
+        sequence += _phonem_to_sequence(phonems_text)
+        sequence.append(_phonemes_to_id[' '])
     sequence.append(_phonemes_to_id['~'])
     return sequence
 
@@ -66,6 +76,17 @@ def sequence_to_text(sequence):
     return result.replace('}{', ' ')
 
 
+def sequence_to_phonem(sequence):
+    '''Converts a sequence of IDs back to a string'''
+    result = ''
+    for symbol_id in sequence:
+        if symbol_id in _id_to_phonemes:
+            s = _id_to_phonemes[symbol_id]
+            print(s)
+            result += s
+    return result.replace('}{', ' ')
+
+
 def _clean_text(text, cleaner_names):
     for name in cleaner_names:
         cleaner = getattr(cleaners, name)
@@ -80,7 +101,7 @@ def _symbols_to_sequence(symbols):
 
 
 def _phonem_to_sequence(phonemes):
-    return [_phonemes_to_id[s] for s in phonemes if _should_keep_phonem(s)]
+    return [_phonemes_to_id[s] for s in phonemes.split(" ") if _should_keep_phonem(s)]
 
 
 def _arpabet_to_sequence(text):

utils/text/cmudict.py

@@ -64,12 +64,11 @@ _phonemes = set(_phonemes)
 
 def text2phone(text):
     seperator = phonemizer.separator.Separator('', '', ' ')
-    ph = phonemizer.phonemize(text, separator=seperator)
-    ph = ph.split(' ')
-    ph.remove('')
-
-    result = [char2code[p] for p in ph]
-    return result
+    try:
+        ph = phonemizer.phonemize(text, separator=seperator)
+    except:
+        ph = None
+    return ph
 
 
 class CMUDict:
@@ -95,6 +94,20 @@ class CMUDict:
         '''Returns list of ARPAbet pronunciations of the given word.'''
         return self._entries.get(word.upper())
 
+    def get_arpabet(self, word, cmudict, punctuation_symbols):
+        first_symbol, last_symbol = '', ''
+        if len(word) > 0 and word[0] in punctuation_symbols:
+            first_symbol = word[0]
+            word = word[1:]
+        if len(word) > 0 and word[-1] in punctuation_symbols:
+            last_symbol = word[-1]
+            word = word[:-1]
+        arpabet = cmudict.lookup(word)
+        if arpabet is not None:
+            return first_symbol + '{%s}' % arpabet[0] + last_symbol
+        else:
+            return first_symbol + word + last_symbol
+
 
 _alt_re = re.compile(r'\([0-9]+\)')
 

utils/text/symbols.py

@@ -10,13 +10,15 @@ from utils.text import cmudict
 _pad = '_'
 _eos = '~'
 _characters = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz!\'(),-.:;? '
+_punctuations = '!\'(),-.:;? '
 
 # Prepend "@" to ARPAbet symbols to ensure uniqueness (some are the same as uppercase letters):
 _arpabet = ['@' + s for s in cmudict._phonemes]
 
 # Export all symbols:
 symbols = [_pad, _eos] + list(_characters) + _arpabet
-phonemes = [_pad, _eos] + cmudict._phonemes + list('!\'(),-.:;?')
+phonemes = [_pad, _eos] + list(cmudict._phonemes) + list(_punctuations)
 
 if __name__ == '__main__':
     print(symbols)
+    print(phonemes)