CurateLM Multilingual Sample
A small sample of the CurateLM corpus: licence-verified multilingual text for AI training. It shows the format and the licence discipline, not the volume. The corpus grows daily.
Contents
| Folder | What it holds |
|---|---|
by_language/ |
127 files, one document excerpt per language |
by_niche/ |
116 files, one document excerpt per subject area (niche) |
Each file is JSON Lines with a single record.
Fields
| Field | Meaning |
|---|---|
text |
the text excerpt |
language |
language code |
niche |
subject area |
license |
licence of this record: PUBLIC-DOMAIN, PUBLIC-DOMAIN-US (public domain in the United States) or CC0-1.0 |
quality_score |
CurateLM quality score (0 to 1) |
token_estimate_sample |
approximate tokens in this excerpt |
token_estimate_full |
approximate tokens in the full source document |
Licensing
Only records verified as public domain or CC0 are included. Material under attribution, share-alike, non-commercial or no-derivatives licences is not part of this sample. Public-domain status can depend on the jurisdiction; the license field states what we verified for each record.
About CurateLM
CurateLM, based in Berlin, licenses multilingual, code and biomedical datasets for AI training and prepares companies' own data for AI use in a GDPR-compliant way. The complete corpus is available under commercial licence by agreement.
Contact: hello@curatelm.com | curatelm.com
- Downloads last month
- 18