JavaScript metaphone
Break english phrases down into their phonemes
1 2 3 4 56 7 8 9 1011 12 13 14 1516 17 18 19 2021 22 23 24 2526 27 28 29 3031 32 33 34 3536 37 38 39 4041 42 43 44 4546 47 48 49 5051 52 53 54 5556 57 58 59 6061 62 63 64 6566 67 68 69 7071 72 73 74 7576 77 78 79 8081 82 83 84 8586 87 88 89 9091 92 93 94 9596 97 98 99 100101 102 103 104 105106 107 108 109 110111 112 113 114 115116 117 118 119 120121 122 123 124 125126 127 128 129 130131 132 133 134 135136 137 138 139 140141 142 143 144 145146 147 148 149 150151 152 153 154 155156 157 158 159 160161 162 163 164 165166 167 168 169 170171 172 173 174 175176 177 178 179 180181 182 183 184 185186 187 188 189 190191 192 193 194 195196 197 198 199 200201 202 203 204 205206 207 208 209 210211 212 213 214 215216 217 218 219 220221 222 223 224 225226 227 228 229 230231 232 233 234 235236 237 238 239 240241 242 243 244 245246 247 | function metaphone (word, phones) { // + original by: Greg Frazier // + improved by: Brett Zamir (http://brett-zamir.me) // + improved by: Rafał Kukawski (http://kukawski.pl) // * example 1: metaphone('Gnu'); // * returns 1: 'N' word = (word == null ? '' : word + '').toUpperCase(); function isVowel (a) { return 'AEIOU'.indexOf(a) !== -1; } function removeDuplicates (word) { var wordlength = word.length, char1 = word.charAt(0), char2, rebuilt = char1; for (var i = 1; i < wordlength; i++) { char2 = word.charAt(i); if (char2 !== char1 || char2 === 'C' || char2 === 'G') { // 'c' and 'g' are exceptions rebuilt += char2; } char1 = char2; } return rebuilt; } word = removeDuplicates(word); var wordlength = word.length, x = 0, metaword = ''; //Special wh- case if (word.substr(0, 2) === 'WH') { // Remove "h" and rebuild the string word = 'W' + word.substr(2); } var cc = word.charAt(0); // current char. Short name cause it's used all over the function var pc = ''; // previous char. There is none when x === 0 var nc = word.charAt(1); // next char var nnc = ''; // 2 characters ahead. Needed later if (1 <= wordlength) { switch (cc) { case 'A': if (nc === 'E') { metaword += 'E'; } else { metaword += 'A'; } x += 1; break; case 'E': case 'I': case 'O': case 'U': metaword += cc; x += 1; break; case 'G': case 'K': case 'P': if (nc === 'N') { x += 1; } break; case 'W': if (nc === 'R') { x += 1; } break; } } for (; x < wordlength; x++) { cc = word.charAt(x); pc = word.charAt(x - 1); nc = word.charAt(x + 1); nnc = word.charAt(x + 2); if (!isVowel(cc)) { switch (cc) { case 'B': if (pc !== 'M') { metaword += 'B'; } break; case 'C': if (x + 1 <= wordlength) { if (word.substr(x - 1, 3) !== 'SCH') { if (x === 0 && (x + 2 <= wordlength) && isVowel(nnc)) { metaword += 'K'; } else { metaword += 'X'; } } else if (word.substr(x + 1, 2) === 'IA') { metaword += 'X'; } else if ('IEY'.indexOf(nc) !== -1) { if (x > 0) { if (pc !== 'S') { metaword += 'S'; } } else { metaword += 'S'; } } else { metaword += 'K'; } } else { metaword += 'K'; } break; case 'D': if (x + 2 <= wordlength && nc === 'G' && 'EIY'.indexOf(nnc) !== -1) { metaword += 'J'; x += 2; } else { metaword += 'T'; } break; case 'F': metaword += 'F'; break; case 'G': if (x < wordlength) { if ((nc === 'N' && x + 1 === wordlength - 1) || (nc === 'N' && nnc === 'S' && x + 2 === wordlength - 1)) { break; } if (word.substr(x + 1, 3) === 'NED' && x + 3 === wordlength - 1) { break; } if (word.substr(x - 2, 3) === 'ING' && x === wordlength - 1) { break; } if (x + 1 <= wordlength - 1 && word.substr(x - 2, 4) === 'OUGH') { metaword += 'F'; break; } if (nc === 'H' && x + 2 <= wordlength) { if (isVowel(nnc)) { metaword += 'K'; } } else if (x + 1 === wordlength) { if (nc !== 'N') { metaword += 'K'; } } else if (x + 3 === wordlength) { if (word.substr(x + 1, 3) !== 'NED') { metaword += 'K'; } } else if (x + 1 <= wordlength) { if ('EIY'.indexOf(nc) !== -1) { if (pc != 'G') { metaword += 'J'; } } else if (x === 0 || pc !== 'D' || 'EIY'.indexOf(nc) === -1) { metaword += 'K'; } } else { metaword += 'K'; } } else { metaword += 'K'; } break; case 'M': case 'J': case 'N': case 'R': case 'L': metaword += cc; break; case 'Q': metaword += 'K'; break; case 'V': metaword += 'F'; break; case 'Z': metaword += 'S'; break; case 'X': metaword += (x === 0) ? 'S' : 'KS'; break; case 'K': if (x === 0 || pc !== 'C') { metaword += 'K'; } break; case 'P': if (x + 1 <= wordlength && nc === 'H') { metaword += 'F'; } else { metaword += 'P'; } break; case 'Y': if (x + 1 > wordlength || isVowel(nc)) { metaword += 'Y'; } break; case 'H': if (x === 0 || 'CSPTG'.indexOf(pc) === -1) { if (isVowel(nc) === true) { metaword += 'H'; } } break; case 'S': if (x + 1 <= wordlength) { if (nc === 'H') { metaword += 'X'; } else if (x + 2 <= wordlength && nc === 'I' && 'AO'.indexOf(nnc) !== -1) { metaword += 'X'; } else { metaword += 'S'; } } else { metaword += 'S'; } break; case 'T': if (x + 1 <= wordlength) { if (nc === 'H') { metaword += '0'; } else if (x + 2 <= wordlength && nc === 'I' && 'AO'.indexOf(nnc) !== -1) { metaword += 'X'; } else { metaword += 'T'; } } else { metaword += 'T'; } break; case 'W': if (x + 1 <= wordlength && isVowel(nc)) { metaword += 'W'; } break; } } } phones = parseInt(phones, 10); if (metaword.length > phones) { return metaword.substr(0, phones); } return metaword; } |
Examples
Running
1 | metaphone('Gnu'); |
Should return
1 | 'N' |
Dependencies
No dependencies, you can use this function standalone.
Open syntax issues
php.js uses JsLint to help us keep our code consistent and prevent some common bugs.
Eventually we want all code to pass or at least take into consideration most fixes suggested by JsLint, following this JsLint configuration we’ve decided on.
Authors
Thanks to the following developers, you get to have metaphone goodness in JavaScript.
@Rey: According to the original BASIC implementation of metaphone at http://aspell.net/metaphone/metaphone.basic (though not mentioned in summaries of the algorithm such as http://en.wikipedia.org/wiki/Metaphone ), "Ch" at the beginning of a word and then followed by a vowel should produce "k". It seems to depend on what metaphone version is being targeted (see http://aspell.net/metaphone/ ). The PHP source code had a "traditional" flag, apparently to distinguish this original algorithm from others, but this was not apparently exposed anywhere in the public API.
Practically speaking, there are examples like "Character", "Charisma", and "Chyme" where "k" makes sense, but other counter-examples like "Check", "Chick", "Chinese", "Chore", etc. which might indicate otherwise (seems 'e', 'i', 'o' should have been distinguished from other vowels based on this sample anyways, though there is also "Chad").
I went with the original algorithm, thinking http://bugs.php.net/bug.php?id=48711 might end up getting PHP to follow the original.
Lines 114-118 in our code deal with this, so feel free to alter it in your own copy, as this seems up for grabs to me (though maybe we should revert to PHP behavior, allowing configuration to allow the deviation until such time as PHP itself may resolve the issue).
@Rey: yeah, the metaphone requires a big cleanup and some bug fixing. But the PHP implementation isn't bug free, nor does it follow the metaphone "specs". There are some differences.
I did a rewrite of the function, that you can see on github https://github.com/kvz/phpjs/raw/master/_workbench/strings/metaphone.js You can check if this one suits your needs (it lacks phonemes param support, but that's easy to do). The problem is, it's difficult to test if a metaphone implementation works correctly. For many cases also the current implementation works, but as you've proven it's not bug free.
I will try to unit test that one ASAP make it a final release.


Lawrence Philips
4 Jun '11
This is in fact an incredibly complicated question, since the number of exceptions in the pronunciation of consonants and consonant groups in English is huge. These issues are only really addressed correctly in the third generation of the Metaphone algorithm, Metaphone 3