It is more complex then just a simple regex e.g.,
"Hi, how are you?" → "Hubi, hubow ubare yubou?"
Simple regex won't catch that e
is not pronounced in are
.
You need a library that provides a pronunciation dictionary such as nltk.corpus.cmudict
:
from nltk.corpus import cmudict # $ pip install nltk
# $ python -c "import nltk; nltk.download('cmudict')"
def spubeak(word, pronunciations=cmudict.dict()):
istitle = word.istitle() # remember, to preserve titlecase
w = word.lower() #note: ignore Unicode case-folding
for syllables in pronunciations.get(w, []):
parts = []
for syl in syllables:
if syl[:1] == syl[1:2]:
syl = syl[1:] # remove duplicate
isvowel = syl[-1].isdigit()
# pronounce the word
parts.append('ub'+syl[:-1] if isvowel else syl)
result = ''.join(map(str.lower, parts))
return result.title() if istitle else result
return word # word not found in the dictionary
Example:
#!/usr/bin/env python
# -*- coding: utf-8 -*-
import re
sent = "Hi, how are you?"
subent = " ".join(["".join(map(spubeak, re.split("(W+)", nonblank)))
for nonblank in sent.split()])
print('"{}" → "{}"'.format(sent, subent))
Output
"Hi, how are you?" → "Hubay, hubaw ubar yubuw?"
Note: It is different from the first example: each word is replaced with its syllables.
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…