Wordfreq 3.1.1, Robyn Speer, https://github.com/rspeer/wordfreq; frequency data CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). Sources include Google Books Ngrams, Leeds Internet Corpus, Wikipedia, ParaCrawl, OPUS OpenSubtitles 2018 / OpenSubtitles, and freely available SUBTLEX data by Marc Brysbaert and colleagues. See app/src/main/assets/licenses/wordfreq-NOTICE.md. Changes: language alphabet and length filters, exclusions, deduplication, uppercase normalization, frequency selection, and combination with original Wortgarten vocabulary. This adapted dictionary is distributed under CC BY-SA 4.0; application code is separate.

Generation V3: the frozen V2 dictionary is additionally filtered using the reviewed exclusions in family-exclusions-v3.json. V1/V2 data remains for compatible saved puzzles. These changes to the adapted dictionary data are distributed under CC BY-SA 4.0. Full license and source information are included here and at https://lokalagent.ai/wortgarten/lizenzen.

Generation V4: new rounds use family-exclusions-v4.json, adding exclusions for explicit substance and violence references and further insults. 85 fixed campaign slots received editorial corrections while preserving their IDs. The current V4 word pool contains 47,618 entries across six languages. Frozen V1–V3 generators remain for compatible saved rounds; affected active rounds are replaced without resetting the round number or earned rewards. The adapted dictionary data remains available under CC BY-SA 4.0. Screening is an editorial aid and does not guarantee an age classification.