Datasets:
Add source: 1000_novels (1000 Novels Corpus, CLARIN-PL, CC BY 4.0)
#1
by bartoszkobylinski1 - opened
Adds the 1000_novels source — the 1000 Novels Corpus (CLARIN-PL), Polish prose.
- License: CC BY 4.0 — verified at the CLARIN-PL source (handle 11321/312). Attribution: Eder, Rybicki, Młynarczyk, Oleksy, Borys, Maryl, Piasecki (Wrocław University of Technology).
- Provenance: pulled from SpeakLeash public redistribution, key
1000_novels_corpus_CLARIN-PL. - Stats: 1,000 docs · 151,610,758 chars · 60,517,363 tokens (tiktoken cl100k proxy).
- created range: 1785–2013 (from per-file edition years). Schema matches repo (
id/text/source/added/created/token_count); minimalbuild_dynaword.pyfilters (read 1000 → kept 1000).
Maintainer note: please run build_dynaword.py to recompute the canonical Llama-3 token_count and update the top-level README aggregates / CHANGELOG / version bump. A reproducibility create.py is available to drop into src/ on request.
kacperwikiel changed pull request status to merged