README.md

metadata

license: bsd-2-clause
task_categories:
  - automatic-speech-recognition
language:
  - pl
pretty_name: LUNA
size_categories:
  - 10K<n<100K
dataset_info:
  features:
    - name: audio
      dtype:
        audio:
          sampling_rate: 16000
    - name: transcript
      dtype: string
    - name: gender
      dtype: string
    - name: id
      dtype: string
  splits:
    - name: train
      num_bytes: 1068204279.3
      num_examples: 10385
  download_size: 1135905315
  dataset_size: 1068204279.3
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

This speech dataset consists of 10385 short audio clips of multiple speakers conversing in Polish. A transcription is provided for each clip, also gender of speaker is provided for part of the dataset. Clips have total length of almost 10 hours.

This dataset was created from LUNA dataset of human-human and human-computer dialogues on the topic of public transport. The postprocessing consisted of extracting segments with human speech together with their transcripts. If you are interested in the original data, please visit https://zil.ipipan.waw.pl/LUNA

Also, if you find this resource helpful, kindly consider leaving a like.