phonemize/dist/en/phonotactics.d.ts

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

/**
 * Universal English phonotactic post-processing.
 *
 * A small, principled set of rules that apply to *any* IPA output
 * regardless of which pipeline produced it. These are pure
 * phonological tweaks (not word-specific patches) that adjust the
 * surface form to match native phonology — happy-tensing, syllabic
 * consonants, etc.
 *
 * Distinct from the retired postBase pile (which was 500+ ad-hoc
 * if/replace rules for specific orthographic patterns). The rules
 * here have ≤10 entries and each one captures a universal pattern.
 *
 * Performance notes:
 *   - Every regex is module-level const so the engine caches its
 *     compilation.
 *   - Each rule is guarded by a cheap substring/endsWith check before
 *     the regex executes. The vast majority of IPA strings don't hit
 *     any given rule, so the guards skip the expensive scan.
 *   - dropSilentH short-circuits when there's no /h/ at all (avoids
 *     allocating a new string).
 */
/**
 * Apply the small phonotactic rule set in the canonical order. Each
 * rule self-guards with a cheap pre-check, so calls that don't trigger
 * any rule do roughly O(rule-count) substring scans + zero allocations.
 *
 * `word` is the lowercased orthographic form, used by rules that need
 * lexical context (e.g. the weak-schwa function-word exemption in
 * normalizeStrut).
 */
export declare function applyPhonotactics(ipa: string, word?: string): string;