Credits
Wortschatz is built from open data, free software and a touch of AI. Here is what we owe to whom.
-
Wortschatz Leipzig
Wordlist source: the German news corpus
deu_news_2024_1M(one million sentences, 2024).Modifications made:
- Lemmatisation via spaCy (model
de_core_news_lg) - Aggregation of frequencies per (lemma, part-of-speech)
- Filtering of digits, punctuation, acronyms and auxiliary verbs
- Segmentation of remaining entries by frequency into CEFR levels A1–C2
- Lemmatisation via spaCy (model
-
Anthropic Claude
The German example sentences and the English translations were generated once during content production via the Claude API and are subsequently shipped as static SQLite data inside the app. During use of the app there is no data exchange with Anthropic.
-
Vollkorn & Rubik
Vollkorn by Friedrich Althausen for the German words and sentences; Rubik by Hubert & Fischer, Meir Sadan and Cyreal for body and UI. All font files are served from the same origin as the rest of the site. No Google Fonts or other font CDNs are involved.
-
App libraries
The app is built on Kotlin Multiplatform and Compose Multiplatform, and uses among others SQLDelight, Koin, AndroidX libraries (Lifecycle, Navigation 3, DataStore, Glance), kotlinx-coroutines, kotlinx-datetime, kotlinx-serialization, and SQLCipher for the encrypted local database.
The full list of bundled libraries with their licence texts is available inside the app under Settings → Credits.
-
Cloudflare Pages
This website is served statically via Cloudflare Pages. Details about the resulting data processing are set out in the privacy policy.
-
Wortschatz source code
The source code of the app is released under the MIT licence. Content (example sentences, translations, website copy and layout) is governed separately — see terms of use, section 4.