Türkçe kelime listesi - kaynak ve lisans
Turkish word list - origin and licence
=======================================

Source:    hunspell-tr 1.1.1 (30 Dec 2023), Turkish Data Depository (tdd-ai): Ali Safaya, Arda Göktoğan,
           Deniz Yuret, Emirhan Kurtuluş, Taner Sezer; https://github.com/tdd-ai/hunspell-tr
           as shipped in the LibreOffice dictionaries, commit 32b006a2c22a4ac7e8ed3f03346f7b3d85a970a4:
           https://raw.githubusercontent.com/LibreOffice/dictionaries/32b006a2c22a4ac7e8ed3f03346f7b3d85a970a4/tr_TR/tr_TR.dic
             SHA-256 48e5352c5770956b24f355265aef7649b25bcb4d933c0b254ea8b419c241f4bd
           .../tr_TR/tr_TR.aff
             SHA-256 a04a227b2ee45574000b876a1afa982e3436d49ebb97a7a796ddc5ec0cc8191b
           .../tr_TR/README.txt (here: README_tr_TR-original.txt), .../tr_TR/LICENSE (here: MPL-2.0.txt)
Retrieved: 2026-09-28
Licence:   Mozilla Public License 2.0 ("This dictionary is licensed under MPL 2.0 License", README.txt).
           MPL-2.0 permits combination with GPL-licensed data (secondary licence, section 3.3).

Processing (scripts/build-wordlist.mjs, function trFormen, run: node scripts/build-wordlist.mjs tr):
- tr_TR.dic is a compressed list of word forms (FLAG num, each flag = exactly one suffix):
  every stem plus every suffix of its flags
- lower-case words of the Turkish alphabet only (a-z without q, w, x, plus ç ğ ı ö ş ü and â î û)
- only forms that occur in wordfreq (tr) - wordfreq 3.1.1 has only the "small" Turkish list,
  so all forms have Zipf >= about 3 (see README_wordfreq.txt)
- proper names removed: (a) bases whose apostrophe forms (Ankara'da, Almanya'nın) reach at least 30 %
  of the frequency of the base word (wordfreq-tr-apostroph.tsv; not for possessive compounds ending
  in -ı/-i/-u/-ü such as üniversitesi), (b) country names from CLDR (Intl.DisplayNames 'tr'),
  (c) a hand-checked list of common first names and surnames (TR_NAMEN: ahmet, ayşe, yılmaz ...;
  names that are also ordinary words such as deniz, aslan, umut, yıldız stay)
- a few slurs removed together with their forms (TR_RAUS)
Result:    53,297 word forms in <length>.txt.
Wordle:    wordle.txt = forms with exactly 5 letters (ç ğ ı ö ş ü are letters of their own,
           ı and i are different letters; â î û count as a i u) -> 4,549 words, inflected forms included.
Limits:    wordfreq does not know upper and lower case; rarer names that are not in TR_NAMEN
           and loan words from other languages can remain.

Licence of the lists in this folder (<length>.txt, wordle.txt): GPL-3.0-or-later
(COPYING-GPL-3.0.txt), because they combine hunspell-tr (MPL-2.0) with wordfreq data (CC BY-SA 4.0).