import re from pypinyin import pinyin, lazy_pinyin, Style with open('data/bpe_train-set.txt', 'w') as out: #aishell3 with open('data/label_train-set.txt', 'r') as f: for i,line in enumerate(f): if i<5: continue text = line.strip().split('|')[2].replace('% ','').replace('$','').replace('%','') pinyin = ' '.join(lazy_pinyin(text, style=Style.TONE3, neutral_tone_with_five=True)) out.write(pinyin+'\n') #data-baker 1w with open('data/000001-010000.txt', 'r') as f: for i,line in enumerate(f): if i%2==1: continue text = line.strip().split('\t')[1] text = re.sub(r'[#\d]', '', text) pinyin = ' '.join(lazy_pinyin(text, style=Style.TONE3, neutral_tone_with_five=True)) out.write(pinyin+'\n')