Merge branch 'develop' of github.com:PaddlePaddle/Parakeet into develop

examples/fastspeech2/baker/frontend.py

@@ -12,6 +12,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import re
+from typing import Dict
+from typing import List
 
 import numpy as np
 import paddle
@@ -35,7 +37,7 @@ class Frontend():
             for tone, id in tone_id:
                 self.vocab_tones[tone] = int(id)
 
-    def _p2id(self, phonemes):
+    def _p2id(self, phonemes: List[str]) -> np.array:
         # replace unk phone with sp
         phonemes = [
             phn if phn in self.vocab_phones else "sp" for phn in phonemes
@@ -43,13 +45,14 @@ class Frontend():
         phone_ids = [self.vocab_phones[item] for item in phonemes]
         return np.array(phone_ids, np.int64)
 
-    def _t2id(self, tones):
+    def _t2id(self, tones: List[str]) -> np.array:
         # replace unk phone with sp
         tones = [tone if tone in self.vocab_tones else "0" for tone in tones]
         tone_ids = [self.vocab_tones[item] for item in tones]
         return np.array(tone_ids, np.int64)
 
-    def _get_phone_tone(self, phonemes, get_tone_ids=False):
+    def _get_phone_tone(self, phonemes: List[str],
+                        get_tone_ids: bool=False) -> List[List[str]]:
         phones = []
         tones = []
         if get_tone_ids and self.vocab_tones:
@@ -88,7 +91,11 @@ class Frontend():
                     phones.append(phone)
         return phones, tones
 
-    def get_input_ids(self, sentence, merge_sentences=True, get_tone_ids=False):
+    def get_input_ids(
+            self,
+            sentence: str,
+            merge_sentences: bool=True,
+            get_tone_ids: bool=False) -> Dict[str, List[paddle.Tensor]]:
         phonemes = self.frontend.get_phonemes(
             sentence, merge_sentences=merge_sentences)
         result = {}

examples/text_frontend/README.md

@@ -15,6 +15,6 @@ Run the command below to get the results of test.
 ```bash
 ./run.sh
 ```
-The `avg WER` of g2p is: 0.027124048652822204
+The `avg WER` of g2p is: 0.027495061517943988
 
-The `avg CER` of text normalization is: 0.0061629764893859846
+The `avg CER` of text normalization is: 0.006391234877881762

parakeet/frontend/cn_frontend.py

@@ -12,6 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import re
+from typing import List
 
 import jieba.posseg as psg
 from g2pM import G2pM
@@ -43,7 +44,7 @@ class Frontend():
             "狗儿"
         }
 
-    def _get_initials_finals(self, word):
+    def _get_initials_finals(self, word: str) -> List[List[str]]:
         initials = []
         finals = []
         if self.g2p_model == "pypinyin":
@@ -78,7 +79,10 @@ class Frontend():
         return initials, finals
 
     # if merge_sentences, merge all sentences into one phone sequence
-    def _g2p(self, sentences, merge_sentences=True, with_erhua=True):
+    def _g2p(self,
+             sentences: List[str],
+             merge_sentences: bool=True,
+             with_erhua: bool=True) -> List[List[str]]:
         segments = sentences
         phones_list = []
         for seg in segments:
@@ -120,7 +124,11 @@ class Frontend():
             phones_list.append(merge_list)
         return phones_list
 
-    def _merge_erhua(self, initials, finals, word, pos):
+    def _merge_erhua(self,
+                     initials: List[str],
+                     finals: List[str],
+                     word: str,
+                     pos: str) -> List[List[str]]:
         if word not in self.must_erhua and (word in self.not_erhua or
                                             pos in {"a", "j", "nr"}):
             return initials, finals
@@ -137,7 +145,10 @@ class Frontend():
                 new_initials.append(initials[i])
         return new_initials, new_finals
 
-    def get_phonemes(self, sentence, merge_sentences=True, with_erhua=True):
+    def get_phonemes(self,
+                     sentence: str,
+                     merge_sentences: bool=True,
+                     with_erhua: bool=True) -> List[List[str]]:
         sentences = self.text_normalizer.normalize(sentence)
         phonemes = self._g2p(
             sentences, merge_sentences=merge_sentences, with_erhua=with_erhua)

parakeet/frontend/cn_normalization/num.py

@@ -28,7 +28,7 @@ UNITS = OrderedDict({
     8: '亿',
 })
 
-COM_QUANTIFIERS = '(匹|张|座|回|场|尾|条|个|首|阙|阵|网|炮|顶|丘|棵|只|支|袭|辆|挑|担|颗|壳|窠|曲|墙|群|腔|砣|座|客|贯|扎|捆|刀|令|打|手|罗|坡|山|岭|江|溪|钟|队|单|双|对|出|口|头|脚|板|跳|枝|件|贴|针|线|管|名|位|身|堂|课|本|页|家|户|层|丝|毫|厘|分|钱|两|斤|担|铢|石|钧|锱|忽|(千|毫|微)克|毫|厘|分|寸|尺|丈|里|寻|常|铺|程|(千|分|厘|毫|微)米|撮|勺|合|升|斗|石|盘|碗|碟|叠|桶|笼|盆|盒|杯|钟|斛|锅|簋|篮|盘|桶|罐|瓶|壶|卮|盏|箩|箱|煲|啖|袋|钵|年|月|日|季|刻|时|周|天|秒|分|旬|纪|岁|世|更|夜|春|夏|秋|冬|代|伏|辈|丸|泡|粒|颗|幢|堆|条|根|支|道|面|片|张|颗|块|元|(亿|千万|百万|万|千|百)|(亿|千万|百万|万|千|百|美|)元|(亿|千万|百万|万|千|百|)块|角|毛|分)'
+COM_QUANTIFIERS = '(朵|匹|张|座|回|场|尾|条|个|首|阙|阵|网|炮|顶|丘|棵|只|支|袭|辆|挑|担|颗|壳|窠|曲|墙|群|腔|砣|座|客|贯|扎|捆|刀|令|打|手|罗|坡|山|岭|江|溪|钟|队|单|双|对|出|口|头|脚|板|跳|枝|件|贴|针|线|管|名|位|身|堂|课|本|页|家|户|层|丝|毫|厘|分|钱|两|斤|担|铢|石|钧|锱|忽|(千|毫|微)克|毫|厘|(公)分|分|寸|尺|丈|里|寻|常|铺|程|(千|分|厘|毫|微)米|米|撮|勺|合|升|斗|石|盘|碗|碟|叠|桶|笼|盆|盒|杯|钟|斛|锅|簋|篮|盘|桶|罐|瓶|壶|卮|盏|箩|箱|煲|啖|袋|钵|年|月|日|季|刻|时|周|天|秒|分|旬|纪|岁|世|更|夜|春|夏|秋|冬|代|伏|辈|丸|泡|粒|颗|幢|堆|条|根|支|道|面|片|张|颗|块|元|(亿|千万|百万|万|千|百)|(亿|千万|百万|万|千|百|美|)元|(亿|千万|百万|万|千|百|)块|角|毛|分)'
 
 # 分数表达式
 RE_FRAC = re.compile(r'(-?)(\d+)/(\d+)')
@@ -74,7 +74,7 @@ def replace_negative_num(match: re.Match) -> str:
 
 # 编号-无符号整形
 # 00078
-RE_DEFAULT_NUM = re.compile(r'\d{4}\d*')
+RE_DEFAULT_NUM = re.compile(r'\d{3}\d*')
 
 
 def replace_default_num(match: re.Match):

parakeet/frontend/cn_normalization/text_normlization.py

@@ -14,6 +14,7 @@
 import re
 from typing import List
 
+from .char_convert import tranditional_to_simplified
 from .chronology import RE_DATE
 from .chronology import RE_DATE2
 from .chronology import RE_TIME
@@ -66,8 +67,9 @@ class TextNormalizer():
         sentences = [sentence.strip() for sentence in re.split(r'\n+', text)]
         return sentences
 
-    def normalize_sentence(self, sentence):
+    def normalize_sentence(self, sentence: str) -> str:
         # basic character conversions
+        sentence = tranditional_to_simplified(sentence)
         sentence = sentence.translate(F2H_ASCII_LETTERS).translate(
             F2H_DIGITS).translate(F2H_SPACE)
 
@@ -90,7 +92,7 @@ class TextNormalizer():
 
         return sentence
 
-    def normalize(self, text):
+    def normalize(self, text: str) -> List[str]:
         sentences = self._split(text)
         sentences = [self.normalize_sentence(sent) for sent in sentences]
         return sentences

parakeet/frontend/tone_sandhi.py

@@ -114,7 +114,6 @@ class ToneSandhi():
                     -2:] in self.must_neural_tone_words:
                 finals_list[i][-1] = finals_list[i][-1][:-1] + "5"
         finals = sum(finals_list, [])
-
         return finals
 
     def _bu_sandhi(self, word: str, finals: List[str]) -> List[str]:
@@ -151,11 +150,9 @@ class ToneSandhi():
                         finals[i] = finals[i][:-1] + "4"
         return finals
 
-    def _split_word(self, word):
+    def _split_word(self, word: str) -> List[str]:
         word_list = jieba.cut_for_search(word)
         word_list = sorted(word_list, key=lambda i: len(i), reverse=False)
-        new_word_list = []
-
         first_subword = word_list[0]
         first_begin_idx = word.find(first_subword)
         if first_begin_idx == 0:
@@ -280,7 +277,7 @@ class ToneSandhi():
 
         return new_seg
 
-    def _is_reduplication(self, word):
+    def _is_reduplication(self, word: str) -> bool:
         return len(word) == 2 and word[0] == word[1]
 
     # the last char of first word and the first char of second word is tone_three