phonemize
Version:
Fast phonemizer with rule-based G2P prediction. Pure JavaScript implementation.
12 lines (11 loc) • 519 B
TypeScript
/**
* Japanese text normalization for preProcess: converts Arabic digits to
* hiragana so the downstream ja-g2p (romaji-based after anyAscii) can read
* them.
*
* Outputs hiragana in base form without rendaku — ja-g2p's syllable map
* lacks the palatal voiced rows (bya/pya/…), so emitting さんびゃく would
* produce broken phonemes for the びゃ slot. We accept さんひゃく etc. as
* the imperfect-but-tokenizable compromise.
*/
export declare function expandJapaneseText(text: string): string;