greek-stemmer-ts

A dependency-free TypeScript implementation of the Ntais (2006) Greek stemmer. Fork of [email protected] by Apmats (2015) — MIT License.

Install

npm install @cadriva/greek-stemmer-ts

Usage

import { stem } from "@cadriva/greek-stemmer-ts";

// Greek input must be pre-normalised: NFD + strip combining marks + uppercase.
const result = stem("ΤΙΜΟΛΟΓ"); // => 'ΤΙΜΟΛΟΓ'

function normaliseGreek(word: string): string {
  return word
    .normalize("NFD")
    .replace(/[̀-ͯ]/g, "")
    .replace(/ς/g, "σ")
    .toUpperCase();
}

console.log(stem(normaliseGreek("τιμολόγια"))); // => 'ΤΙΜΟΛΟΓ'
console.log(stem("invoice")); // => 'invoice' (non-Greek passes through)
console.log(stem("")); // => ''

Notes

• Operates on monotonic uppercase Greek text only.
• Polytonic Greek is not supported — normalise polytonic input to monotonic before calling stem().
• No language detection, tokenisation, or normalisation: callers pre-normalise. Non-Greek inputs (Latin, numbers, etc.) pass through unchanged.

Algorithm

Ntais (2006) Greek stemmer, implemented as a Snowball-style cascade of regex-based suffix-stripping rules. Pure TypeScript, zero runtime dependencies, zero network/disk/eval surface.

Two rule differences from [email protected] upstream:

| Step | Rule | Upstream | This package | | ------- | -------------------------------- | -------------------------------------------------- | ------------------------------- | | Step 2a | ΑΔΕΣ/ΑΔΩΝ exception list | includes ΓΑΛΑΤ, ΦΑΦΛΑΤ | does not include them | | Step 5h | ΟΥΣΑ/ΟΥΣΕΣ/ΟΥΣΕ retention list | missing ΔΕ, ΔΕΥΤΕΡΕΥ, ΚΑΘΑΡΕΥ, ΠΛΕ, ΤΣΑ | includes all five |

See src/index.ts comments for source citations.

License

MIT — see LICENSE.