Observed arrival · 2026-11-21
Wieszcz-XIX: Polish language models from 1800–1918
An independent research project publishes a 3.1-billion-word corpus of historical Polish and language models trained exclusively on it.
- For
- Researchers in historical Polish language and literature
- Worth noticing
- The project reports 90 definitely later texts in the Wolne Lektury subset—0.04% of the corpus.
Field notes
The corpus draws on library scans and transcriptions, with source and rights information recorded per document. Its temporal boundary uses catalog dates followed by content checks; the project reports 90 definitely post-1918 texts among the Wolne Lektury materials, or 0.04% of the corpus. The page also flags OCR errors and inherited period prejudices, while its model comparison holds architecture, training steps, and seed constant as parameter count changes.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue