phonemize/dist/en/reduce.d.ts

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

/**
 * English vowel reduction (P2.2 of G2P redesign).
 *
 * Given an IPA string (no stress marks) and the set of stressed nucleus
 * indices, replace each unstressed vowel nucleus with its reduced form.
 *
 * Reduction rules (simple, ordered by specificity):
 *
 *   1. Unstressed nucleus immediately followed by /ɹ/ (any vowel + r in
 *      the same syllable) → ɝ.  Examples: doctor /dɑkˈtɝ/, mother /mʌðˈɝ/.
 *
 *   2. Unstressed word-final open syllable with high front vowel
 *      (/i, ɪ/) → i (happy-tensing).
 *
 *   3. All other unstressed nuclei → ə.
 *
 * Stressed nuclei are left untouched. Long vowels (/i/, /u/, /eɪ/, /oʊ/,
 * /aɪ/, /aʊ/, /ɔɪ/) generally don't reduce — caller can decide whether to
 * apply rules to them via the `reduceLong` option (default: false).
 *
 * The function expects no stress marks in the input — caller manages
 * stress externally via the indices.
 */
/**
 * Locate each vowel nucleus in `ipa`. Returns array of (startIdx, endIdx)
 * where endIdx is exclusive.
 */
export declare function findNuclei(ipa: string): Array<[number, number]>;
export interface ReduceOptions {
    /** Also reduce long vowels (default: false — they typically resist). */
    reduceLong?: boolean;
    /** Apply happy-tensing for unstressed word-final /i/. */
    happyTensing?: boolean;
}
/**
 * Reduce unstressed vowels in `ipa`. `stressed` is a set of nucleus indices
 * (0-based, in the order they appear in `ipa`) that should NOT be reduced.
 */
export declare function reduceUnstressedVowels(ipa: string, stressed: ReadonlySet<number>, opts?: ReduceOptions): string;
/**
 * Given an IPA string with stress marks, extract the set of stressed
 * nucleus indices (primary + secondary both count as "stressed" for
 * reduction purposes).
 */
export declare function extractStressedNuclei(ipa: string): Set<number>;