@ingglish/g2p 0.1.0

package/dist/index.d.cts

@@ -0,0 +1,81 @@
+import { OutputFormat } from '@ingglish/phonemes';
+
+/**
+ * NRL/Elovitz letter-to-sound rules for grapheme-to-phoneme conversion.
+ *
+ * Based on: Elovitz, H., Johnson, R., McHugh, A., and Shore, J. (1976)
+ * "Automatic translation of English text to phonetics by means of
+ * letter-to-sound rules." NRL Report 7948.
+ *
+ * ~960 context-sensitive rules that examine left and right letter context
+ * to determine pronunciation. ~150 are general spelling rules (silent E,
+ * digraphs, vowel teams, C/G softening, suffixes). The other ~800 are
+ * narrow context patches and whole-word memorizations for the cases where
+ * those rules break down.
+ *
+ * Each rule has the form:
+ *
+ *   leftContext[TARGET]rightContext=/PHONEMES/
+ *
+ * Special context symbols:
+ *   #  = 1+ vowels (AEIOUY)
+ *   .  = voiced consonant (BDVGJLMNRWZ)
+ *   %  = suffix (ER, E, ES, ED, ING, ELY)
+ *   &  = sibilant (S, C, G, Z, X, J, CH, SH)
+ *   @  = non-palate consonant (T, S, R, D, L, Z, N, J, TH, CH, SH)
+ *   ^  = single consonant
+ *   +  = front vowel (E, I, Y)
+ *   :  = 0+ consonants
+ *   ' '= word boundary (space)
+ */
+
+interface G2PTrace {
+    phonemes: string[];
+    steps: G2PTraceStep[];
+}
+interface G2PTraceStep {
+    letters: string;
+    phonemes: string[];
+    rule: string;
+}
+/**
+ * Converts a word to ARPAbet using NRL context-sensitive rules.
+ *
+ * @param word The word to convert
+ * @returns Array of ARPAbet phonemes
+ */
+declare function wordToArpabet(word: string): string[];
+/**
+ * Converts a word to ARPAbet using NRL context-sensitive rules,
+ * returning both the phonemes and a trace of which rules fired.
+ */
+declare function wordToArpabetTraced(word: string): G2PTrace;
+/**
+ * Translates an unknown word using NRL letter-to-sound rules.
+ * This is a fallback when the word isn't in the dictionary.
+ *
+ * @param word The unknown word
+ * @param format The output format
+ * @returns The best-effort translation
+ */
+declare function wordToPhonetic(word: string, format?: OutputFormat): string;
+
+/**
+ * Stress prediction post-processor for G2P rules.
+ *
+ * The NRL rules assign stress-1 to ALL non-schwa vowels, meaning multi-syllable
+ * words get every vowel stressed. This module predicts which syllable gets
+ * primary stress using Rastle & Coltheart-style affix rules, then reduces
+ * unstressed vowels to schwa where appropriate.
+ */
+/**
+ * Apply stress prediction to a phoneme array produced by NRL rules.
+ *
+ * - Finds all vowel positions (phonemes ending in 0/1/2)
+ * - Predicts which syllable gets primary stress
+ * - Reduces unstressed reducible vowels to AH0 (schwa)
+ * - Marks other unstressed vowels with 0
+ */
+declare function applyStressPrediction(word: string, phonemes: string[]): string[];
+
+export { type G2PTrace, type G2PTraceStep, applyStressPrediction, wordToArpabet, wordToArpabetTraced, wordToPhonetic };

package/dist/index.d.ts

@@ -0,0 +1,81 @@
+import { OutputFormat } from '@ingglish/phonemes';
+
+/**
+ * NRL/Elovitz letter-to-sound rules for grapheme-to-phoneme conversion.
+ *
+ * Based on: Elovitz, H., Johnson, R., McHugh, A., and Shore, J. (1976)
+ * "Automatic translation of English text to phonetics by means of
+ * letter-to-sound rules." NRL Report 7948.
+ *
+ * ~960 context-sensitive rules that examine left and right letter context
+ * to determine pronunciation. ~150 are general spelling rules (silent E,
+ * digraphs, vowel teams, C/G softening, suffixes). The other ~800 are
+ * narrow context patches and whole-word memorizations for the cases where
+ * those rules break down.
+ *
+ * Each rule has the form:
+ *
+ *   leftContext[TARGET]rightContext=/PHONEMES/
+ *
+ * Special context symbols:
+ *   #  = 1+ vowels (AEIOUY)
+ *   .  = voiced consonant (BDVGJLMNRWZ)
+ *   %  = suffix (ER, E, ES, ED, ING, ELY)
+ *   &  = sibilant (S, C, G, Z, X, J, CH, SH)
+ *   @  = non-palate consonant (T, S, R, D, L, Z, N, J, TH, CH, SH)
+ *   ^  = single consonant
+ *   +  = front vowel (E, I, Y)
+ *   :  = 0+ consonants
+ *   ' '= word boundary (space)
+ */
+
+interface G2PTrace {
+    phonemes: string[];
+    steps: G2PTraceStep[];
+}
+interface G2PTraceStep {
+    letters: string;
+    phonemes: string[];
+    rule: string;
+}
+/**
+ * Converts a word to ARPAbet using NRL context-sensitive rules.
+ *
+ * @param word The word to convert
+ * @returns Array of ARPAbet phonemes
+ */
+declare function wordToArpabet(word: string): string[];
+/**
+ * Converts a word to ARPAbet using NRL context-sensitive rules,
+ * returning both the phonemes and a trace of which rules fired.
+ */
+declare function wordToArpabetTraced(word: string): G2PTrace;
+/**
+ * Translates an unknown word using NRL letter-to-sound rules.
+ * This is a fallback when the word isn't in the dictionary.
+ *
+ * @param word The unknown word
+ * @param format The output format
+ * @returns The best-effort translation
+ */
+declare function wordToPhonetic(word: string, format?: OutputFormat): string;
+
+/**
+ * Stress prediction post-processor for G2P rules.
+ *
+ * The NRL rules assign stress-1 to ALL non-schwa vowels, meaning multi-syllable
+ * words get every vowel stressed. This module predicts which syllable gets
+ * primary stress using Rastle & Coltheart-style affix rules, then reduces
+ * unstressed vowels to schwa where appropriate.
+ */
+/**
+ * Apply stress prediction to a phoneme array produced by NRL rules.
+ *
+ * - Finds all vowel positions (phonemes ending in 0/1/2)
+ * - Predicts which syllable gets primary stress
+ * - Reduces unstressed reducible vowels to AH0 (schwa)
+ * - Marks other unstressed vowels with 0
+ */
+declare function applyStressPrediction(word: string, phonemes: string[]): string[];
+
+export { type G2PTrace, type G2PTraceStep, applyStressPrediction, wordToArpabet, wordToArpabetTraced, wordToPhonetic };