NYSIIS encoder

NYSIIS—the New York State Identification and Intelligence System phonetic code—was designed in 1970 to index personal and family names with higher match accuracy than classic American Soundex. It keeps an alphabetic key that preserves more of the original spelling shape while still folding vowels and common consonant groups so homophones and near-homophones land on the same code. This nysiis encoder accepts a non-empty name string, strips non-letters, applies the standard prefix and suffix translations, scans the remaining letters with the official collapse rules, and returns the resulting code together with the normalized letter form used for coding.

How NYSIIS builds a phonetic key from a name

NYSIIS starts by uppercasing the name and keeping only Latin letters A through Z; digits, spaces, hyphens, apostrophes, and punctuation are stripped so a form field with O'Brien or brown, sr still yields a stable letter stream after cleaning. Suffix rules rewrite the end: EE or IE becomes Y, and DT, RT, RD, NT, or ND becomes D. The first letter of that transformed string is kept as the first character of the key.

From the second letter onward the scan applies one rule at a time: EV becomes AF; any other vowel becomes A; Q becomes G; Z becomes S; M becomes N; KN becomes N and other K becomes C; SCH becomes SSS and PH becomes FF; H is replaced by the previous letter when either neighbor is not a vowel; W after a vowel is replaced by that previous letter. Adjacent identical letters are collapsed so runs never repeat.

After the scan finishes, trailing S is removed, a trailing AY pair becomes Y, and a trailing A is removed.

What to enter and the mistakes to avoid

On success the response includes code (the NYSIIS alphabetic value), method equal to nysiis so downstream routers can branch without sniffing shapes, name echoing the original string, normalized with the uppercase A–Z sequence actually coded, empty always false on success, and length equal to the original character count before stripping. Do not treat NYSIIS as encryption or as a unique person identifier: many unrelated surnames can still collide, and that controlled collision is the feature for fuzzy match, not a bug.

Prefer this landing when search intent is nysiis encoder, New York State phonetic code, or name matching beyond Soundex; American Soundex, NATO phonetic spelling, and Metaphone answer different intents and stay on separate pages. Accented letters outside pure A–Z are stripped with other non-letters, so feed an ASCII-normalized form if your source data uses diacritics and you need those letters to participate. This implementation returns the full code after trailing cleanup rather than forcing a six-character truncate, so longer surnames keep the extra detail NYSIIS was designed to retain.

Where NYSIIS name matching helps

Genealogists and local-history projects use a nysiis encoder to group census, court, and passenger-list surnames that were recorded with inconsistent spelling while retaining more shape than a four-character Soundex bucket. CRM and customer-support tooling can pre-cluster look-alike last names before a human review step without standing up a full entity-resolution engine. Classroom demos of historical indexing algorithms need a deterministic oracle students can call from a browser while the teacher walks through prefix and vowel rules on the board.

Data-cleaning notebooks can emit NYSIIS as one feature column next to lowercase and Soundex columns when building entity-resolution features for record linkage. Pair with other text utilities only when a later step needs a different phonetic or cipher scheme; keep NYSIIS on its own landing so SEO intent stays sharp and the guide remains honest about what the alphabetic code can and cannot do compared with Soundex.