Danish word lists (words-3.txt, words-4.txt, words-5.txt in this folder)

Source: COR1 version 1.5.1.0, Det Centrale Ordregister, by Dansk Sprognævn.
        https://ordregister.dk/files/cor1.5.1.0.tsv  (UTF-8 TSV, 31,206,228 bytes,
        sha256 1c4d1c06bd676e66be8e8c0af68615c8a892472a05af3b261ba8a9d1f2d2b82b)
        Published 16.12.2025; corresponds to Retskrivningsordbogen 5.1.

Licence, as ordregister.dk states it for the COR1 grundressource
(https://ordregister.dk/doc/Gamle-udgaver.html, last line of the COR1 section):

    "Licens: CC0"

and Dansk Sprognævn (https://dsn.dk/forskning/sprogteknologi-og-fagsprog/cor/):

    "Det Centrale Ordregister er frit tilgængeligt for alle og må anvendes til alle formål."

CC0 asks for no attribution. We give it anyway because it is true: the spelling,
the word forms and their inflections all come from Dansk Sprognævn's work.
Thank you, Dansk Sprognævn.

What dankpuzzles did to it (tools/build-words-da.cjs): kept only forms marked N
(listed in Retskrivningsordbogen), written with a-z, æ, ø, å only, whose lemma is
lowercase and which are not abbreviations (fork) or fixed-phrase-only words
(iflerord); kept 3-, 4- and 5-letter forms; removed slurs (tools/blocklist-da.cjs).

The four tiers (#1 common, #2 everyday, #3 less common, #4 rare) are NOT from COR.
They come from word counts over three CC0 subsets of the Danish Dynaword corpus
(https://huggingface.co/datasets/danish-foundation-models/danish-dynaword):
ft (Folketinget debates), danske-taler (speeches) and nordjyllandnews (TV2 Nord
articles), each marked "license: cc0-1.0" on its dataset card. Only a per-million
rate per word was derived (tools/data/da-freq.tsv); no corpus text is shipped.
A few tiers were set by hand (tools/data/da-tier-overrides.txt), and plain
profanity is always tier 4.