schneiderkamplab/opus-da-en-permissive

OPUS da-en permissive subset

This dataset is a processed Danish-English sentence-pair subset harvested from OPUS corpora available at https://opus.nlpl.eu/.

The subset was selected from OPUS da-en sources with permissive Creative Commons or public-domain style licensing, then converted into JSONL.GZ for DFM/HRM-Text training. It is intended to preserve the exact rows used by the local DFM data pipeline, so DFM8 can be rebuilt without relying on local-only files.

Files

• data/opus_da_en.jsonl.gz

Schema

Each row is a JSON object:

{"id": "OPUS-DGT-0", "da": "Traktat", "en": "Treaty", "source": "DGT"}

Fields:

• id: stable local row id.
• da: Danish sentence/text.
• en: English sentence/text.
• source: OPUS source corpus identifier.

DFM usage

In the HRM-Text DFM8 pipeline this dataset corresponds to the local source group opus and the sampling prefix opus__.

The local tokenized DFM8 inventory for this source was approximately 5.66B rendered/source tokens. The DFM7 sampled mix used approximately 14.52B sampled tokens from this source over five epochs.

License and provenance

OPUS aggregates many corpora with different licenses. This processed subset was constructed from OPUS da-en sources selected for permissive Creative Commons or public-domain style licensing. Users should consult OPUS and the original OPUS source-corpus metadata for source-specific license details.