Datasetpublic

NRK-KIHUB/no-asr-eval-data-broadcast-sample

NRK Norwegian Speech Dataset (Sample) Dataset Description Note: This is a sample dataset containing a subset of chunks for demonstration and preview purposes. The full dataset is available privately. This dataset contains Norwegian speech data from NRK TV broadcasts (norge-rundt, supernytt), processed for automatic speech recognition (ASR) evaluation and research. Reference text caveat: Reference text is NRK's on-air teletext subtitling, not a verbatim… See the full description on the dataset page: https://huggingface.co/datasets/NRK-KIHUB/no-asr-eval-data-broadcast-sample.

sourceHugging Faceupdated 2d agoView on Hugging Face

0likes81downloads

Dataset Card

NRK Norwegian Speech Dataset (Sample)

Dataset Description

Note: This is a sample dataset containing a subset of chunks for demonstration and preview purposes. The full dataset is available privately.

This dataset contains Norwegian speech data from NRK TV broadcasts (norge-rundt, supernytt), processed for automatic speech recognition (ASR) evaluation and research.

Reference text caveat: Reference text is NRK's on-air teletext subtitling, not a verbatim transcription made for this dataset. Human-QA'd for broadcast, but condensed for reading speed.

Dataset Statistics

—Total chunks: 108
—Episodes: 36
—Total duration: 0.15 hours
—Chunk types: subtitle_aligned
—Transcription sources: NRK broadcast subtitles (teletext)
—Sample rate: 16,000 Hz
—Language: Norwegian (Bokmål/Nynorsk)

Splits

—test: 108 chunks

Episodes

—DVNR04003625
—DVNR04003725
—DVNR04003825
—DVNR04003925
—DVNR04004025
—DVNR04004125
—DVNR04004225
—DVNR04004325
—DVNR04004425
—DVNR04004525
—DVNR04004625
—DVNR04004725
—DVNR04004825
—DVNR04004925
—DVNR04005025
—DVNR04005125
—DVNR04005225
—DVNR04005325
—IBOT25050426
—IBOT25050526
—IBOT25050626
—IBOT25050726
—IBOT25050826
—IBOT25051126
—IBOT25051226
—IBOT25051326
—IBOT25051526
—IBOT25051826
—IBOT25051926
—IBOT25052026
—IBOT25052126
—IBOT25052226
—IBOT25052626
—IBOT25052726
—IBOT25052826
—IBOT25052926

Dataset Structure

Each data point contains:

—id: Unique chunk identifier
—audio: Audio data (WAV format, 16kHz)
—text: Original transcription text
—text_normalized: Normalized text (lowercase, standardized)
—duration_seconds: Audio duration
—chunk_type: Type of chunk (subtitlealigned or fixedduration)
—episode_id: Source episode identifier
—program_name: NRK program name
—start_time: Start time in source video (seconds)
—end_time: End time in source video (seconds)

Source

—Original videos: NRK TV (https://tv.nrk.no)
—Transcriptions: NRK broadcast subtitles (teletext)
—Processing: Automated pipeline with yt-dlp and custom chunking

Usage

python

from datasets import load_dataset

# Load dataset
dataset = load_dataset("NRK-KIHUB/no-asr-eval-data-broadcast-sample")

# Access first example
example = dataset["test"][0]
print(f"Text: {{example['text']}}")
print(f"Duration: {{example['duration_seconds']:.2f}}s")

# Play audio (in Jupyter/Colab)
from IPython.display import Audio
Audio(example['audio']['array'], rate=example['audio']['sampling_rate'])

License

TBD - Please verify NRK terms of use before distribution.

Citation

If you use this dataset, please cite:

@misc{nrk_norwegian_speech_sample,
  title={NRK Norwegian Speech Dataset (Sample)},
  author={NRK KI-hub},
  year={2026},
  url={https://huggingface.co/datasets/NRK-KIHUB/no-asr-eval-data-broadcast-sample}
}

Contact

For questions or issues, please open an issue on the dataset repository.