syvai/p1-segments

DR P1 speech segments

Dataset

Danish speech clips from DR P1, in mono 16 kHz OGG/Opus, with verbatim text, timing, and speaker metadata. Transcript text and speaker attribution may contain automated errors.

Source

The recordings cover roughly 2006–2022 and come from DR P1 recordings in kb.dk’s DR archive. Audio is sourced through the pinned syvai/p1 revision 449b9c2294026df6d0d37538f279fdec03f565ff. Transcripts were generated with ElevenLabs Scribe v2 through the pinned syvai/p1-transcripts revision 41132579816d86e889635f84f30511279f026359.

Access

This dataset is publicly accessible. Load it with an immutable dataset commit revision (never main):

from datasets import load_dataset

dataset = load_dataset(
    "syvai/p1-segments",
    # Legacy-only readers used data_files="data/train/*.parquet".
    data_files=[
        "data/train/**/*.parquet",
        "data-shards/train/**/*.parquet",
    ],
    revision="<immutable-commit-sha>",
    streaming=True,
)

Licence

The dataset uses a layered licence. The dataset structure, arrangement, segment boundaries, and independently created metadata are CC BY 4.0; embedded DR audio and verbatim/source-derived transcript text are excluded. Public availability does not place excluded content under an open licence or independently grant downstream rights. Users need relevant permissions or another basis under applicable law. See LICENSE.