Rijgersberg/YouTube-Commons-nl-transcriptions

YouTube Commons NL transcriptions

This is the Rijgersberg/YouTube-Commons-descriptions dataset, filtered for Dutch and with the Dutch transcriptions added from YouTube-Commons. All of these fall under a CC BY 4.0 license.

Source

The original source of the dataset (minus the titles, descriptions and audio files) is YouTube Commons:

YouTube-Commons is a collection of audio transcripts of 2,063,066 videos shared on YouTube under a CC BY 4.0 license.

Content

The collection comprises 22,709,724 original and automatically translated transcripts from 3,156,703 videos (721,136 individual channels).

Rijgersberg/YouTube-Commons-descriptions added titles, descriptions and language detection on top of that. This dataset adds transcriptions marked as Dutch from the original YouTube-Commons dataset.

Code

The dataset was created using the following code:

from datasets import load_dataset


yt_descriptions = load_dataset('Rijgersberg/YouTube-Commons-descriptions', split='train')
yt_descriptions = yt_descriptions.filter(lambda lang: lang == 'nl', input_columns='language')

nl_videos = set(yt_descriptions['video_id'])

yt_commons = load_dataset('Rijgersberg/YouTube-Commons', split='train')
yt_commons = yt_commons.filter(lambda video_id: video_id in nl_videos, input_columns='video_id')

nl_transcriptions = {
    row['video_id']: row['text']
    for row in yt_commons
    if row['transcription_language'] == 'nl'
}

yt_descriptions = yt_descriptions.map(lambda row: {'transcription': nl_transcriptions.get(row['video_id'])})

yt_descriptions.push_to_hub('Rijgersberg/YouTube-Commons-nl-transcriptions')