blackteay/GPT-SoVITS
1
0
Fork 0
mirror of https://github.com/RVC-Boss/GPT-SoVITS.git synced 2025-10-09 00:10:00 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix homograph dict

Browse Source
This commit is contained in:
KamioRinn 2024-03-21 15:42:09 +08:00
parent 97f304c2bf
commit fdcd409ea8
1 changed files with 7 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
GPT_SoVITS/text/english.py
View File

@ -234,6 +234,10 @@ class en_G2p(G2p):
for word in ["AE", "AI", "AR", "IOS", "HUD", "OS"]: for word in ["AE", "AI", "AR", "IOS", "HUD", "OS"]:
del self.cmu[word.lower()] del self.cmu[word.lower()]
# 修正多音字
self.homograph2features["read"] = (['R', 'IY1', 'D'], ['R', 'EH1', 'D'], 'VBP')
self.homograph2features["complex"] = (['K', 'AH0', 'M', 'P', 'L', 'EH1', 'K', 'S'], ['K', 'AA1', 'M', 'P', 'L', 'EH0', 'K', 'S'], 'JJ')
def __call__(self, text): def __call__(self, text):
# tokenization # tokenization
@ -260,6 +264,9 @@ class en_G2p(G2p):
pron1, pron2, pos1 = self.homograph2features[word] pron1, pron2, pos1 = self.homograph2features[word]
if pos.startswith(pos1): if pos.startswith(pos1):
pron = pron1 pron = pron1
# pos1比pos长仅出现在read
elif len(pos) < len(pos1) and pos == pos1[:len(pos)]:
pron = pron1
else: else:
pron = pron2 pron = pron2
else: else:
@ -306,7 +313,6 @@ class en_G2p(G2p):
# 尝试进行分词，应对复合词 # 尝试进行分词，应对复合词
comps = wordsegment.segment(word.lower()) comps = wordsegment.segment(word.lower())
# 无法分词的送回去预测 # 无法分词的送回去预测
if len(comps)==1: if len(comps)==1:
return self.predict(word) return self.predict(word)