phonemize/dist/ru/g2p.d.ts

Fast phonemizer with rule-based G2P prediction. Pure JavaScript implementation.

import { LanguageProcessor } from "../g2p";
declare class RussianG2P implements LanguageProcessor {
    readonly id = "ru-g2p";
    readonly name = "Russian G2P Processor";
    readonly supportedLanguages: string[];
    preProcess(text: string): string;
    predict(word: string, language?: string, pos?: string): string | null;
    /**
     * Token-level G2P. Input is the post-anyAscii Latin transliteration:
     *
     * - digraphs (zh/kh/ch/sh/shch/ts) → real IPA, looked up longest-first
     *   so the bare `s`/`z`/`t`/`c`/`k`/`h` fallbacks never preempt them.
     * - "y" before a/o/u is the anyAscii rendering of я/ё/ю — when it
     *   follows a consonant it palatalizes that consonant and the `y` is
     *   absorbed; word-initially or after a vowel it surfaces as /j/.
     * - "e" and "i" after a consonant also palatalize.
     * - Final voiced obstruent → voiceless (final devoicing).
     *
     * This is a phonemic approximation only — Russian needs lexical stress
     * to do unstressed-vowel reduction (akan'e / ikan'e) properly, and we
     * don't have a stress dictionary, so vowel quality stays full.
     */
    private processRussian;
    /**
     * Assign heuristic stress and apply unstressed-vowel reduction in place.
     * Vowel tokens are a/e/i/o/u/ɨ. Stressed vowel keeps full quality; for
     * unstressed: o,a → ɐ (first-pretonic or word-initial) else ə; e → ɪ;
     * i/u/ɨ stay. A ˈ mark is inserted at the stressed syllable's onset.
     */
    private applyStressAndReduction;
    addPronunciation(word: string, pronunciation: string): void;
}
export default RussianG2P;