What the shared cache contains
The cache contains standard language facts such as a German dictionary form, part of speech, article, inflected forms, an English meaning, and a reusable study note. It does not contain learner identities, email addresses, source pages, URLs, or surrounding private context.
Quick dictionary meanings
Free dictionary meanings come from Wiktionary contributors, as extracted by kaikki.org (wiktextract by Tatu Ylonen, September 2026 extract), and from WikDict by Karl Bartel (Wiktionary data via DBnary, June 2026 release). Linota keeps single German words with their English meanings, articles, and inflected forms, so a selected form such as “ging” finds “gehen”. Unknown compounds are split into known parts, shown as a compound, and a word one letter away can be suggested when you look a word up. The adapted datasets are available under CC BY-SA 4.0: download the combined dataset or the WikDict extract, each with its source checksum. Dictionary meanings are possible meanings, not guaranteed contextual translations. These meanings do not require an AI request; the original dictionaries and their contributors do not endorse Linota.
WordHoard morphology and frequency data
Core German lemmas, frequency ranks, articles, and labeled forms are derived from the WordHoard German frequency dictionary, version 0.1.0. WordHoard is licensed under CC BY-SA 4.0 and includes upstream attribution in its NOTICE file. Linota records the exact source version and checksum with each import.
Generated learning notes
English explanations, examples, and review-card text may be generated with OpenAI models from verified dictionary facts. Generated notes pass structural checks, source-backed grammar reconciliation, an independent quality gate, and staged review before publication. Model output never replaces trusted article or morphology fields.
Corrections
Language is nuanced and errors are still possible. Report a questionable entry with Report a mistake under the study note or reading in the app, or write to help@linota.de. Published notes are versioned so a correction can supersede an old note without erasing its audit trail.
Reading room
Reading room uses original German texts prepared in advance with the help of AI language models from OpenAI and Anthropic and spot-checked by Linota. The everyday explainers are learning material rather than news or professional advice. The texts undergo automated structure and language checks, but errors can remain. No learner data is used to write them. Opening the daily reading does not trigger a new AI request. The separate Your Next Read feature links to external sources with their own attribution.