What this tool does
Metaphone is a phonetic algorithm by Lawrence Philips (1990) that produces a short key representing how an English word sounds. Unlike Soundex, which maps letters with a flat table, Metaphone applies English pronunciation rules to letter combinations, so it produces more accurate matches with fewer collisions. This encoder computes the key in your browser using the 16 standard Metaphone consonant symbols.
How it works
The algorithm first handles silent initial clusters: leading GN, KN, PN, WR and AE drop their first letter, initial X becomes S, and initial WH becomes W. It then walks the word left to right, emitting consonant symbols according to context. Key transformations include:
PH → F (phone → FN)
TH → 0 (think → 0NK, where 0 represents the "th" sound)
SH / -SIO → X (ship → XP)
CH → X (chair → XR)
silent GH → dropped (night → NT)
C before I/E/Y → S; otherwise K
Vowels are retained only at the start of the word and dropped elsewhere, doubled consonants are collapsed, and the 16-symbol alphabet (B X S K J T F H L M N P R 0 W Y) keeps the key compact. The result is a phonetic fingerprint: differently spelled but similar-sounding words converge on the same key.
Practical examples
| Input | Metaphone key | Notes |
|---|---|---|
| Knight | NT | K and GH are both silent |
| Phone | FN | PH → F; vowels after first position dropped |
| Thompson | 0MPSN | TH → 0; silent P in “mps” |
| Catherine | K0RN | C → K; TH → 0; vowels stripped |
| Smith | SM0 | TH → 0 |
Notice that “Catherine” and “Kathryn” both encode to K0RN — the algorithm correctly identifies them as phonetically equivalent despite very different spellings. This is the core value for name-matching problems.
Where Metaphone is most useful
Deduplication in CRM or contact databases — when the same person appears as “Johnston”, “Johnstone”, and “Jonston”, all three encode to the same key, flagging them as candidates for merging. Always follow up with Levenshtein distance or manual review to avoid false positives.
Search recall expansion — a user who types “Smythe” when the database stores “Smith” will miss the record in an exact-match search but both share the key SM0. Generate Metaphone keys at index time and at query time to catch these misses.
Legacy data cleanup — handwritten or OCR-scanned data often produces phonetically correct but orthographically wrong names. Metaphone is particularly effective for English surname normalization in historical records and genealogy databases.
Soundex vs. Metaphone: when to choose which
Soundex is computationally trivial and built into many SQL databases (SOUNDEX() in MySQL, PostgreSQL, and SQL Server), making it the default for simple pipelines. Metaphone is better for English when accuracy matters: it handles digraphs, silent letters, and vowel-initial words more reliably. For names with genuinely ambiguous pronunciations — “Kurzweil” vs. “Kurzwile”, or names from non-English traditions — the Double Metaphone encoder returns a primary and an alternate key, which widens recall further. All processing runs locally in your browser.