chinese_text_normalization

Why we included this project

Chinese ASR and TTS work eventually runs into the gap between raw transcripts and the clean text that lexicons and downstream systems expect. This project closes that gap with a ready-to-run normalization module: it converts numbers, dates, fractions, money, percentages, and phone numbers into their spoken Chinese forms, and strips punctuation in both Chinese and English. It reads plain text, Kaldi archives, and TSV files, so it drops into existing Kaldi-style pipelines with little setup. The normalizers are rule-based and editable, which helps when you need to trace why a particular utterance was rewritten. If you want a working starting point for Chinese text normalization rather than a general NLP toolkit, this is a solid, self-contained reference.