Hugging Face's logo

Datasets:
talkbank
/
callhome

Modalities:
Audio
Text
Formats:
parquet
Languages:
English
Japanese
Chinese
Size:
< 1K
Tags:
speaker-diarization
speaker-segmentation
voice-activity-detection
Libraries:
Datasets
Dask
License:
Dataset card Viewer Files Community
4
callhome / README.md
kamilakesbi's picture
kamilakesbi HF staff
Update README.md
17c8a15 verified
preview code
|
raw
history blame contribute delete
No virus
5.12 kB
metadata
dataset_info:
  - config_name: deu
    features:
      - name: audio
        dtype:
          audio:
            sampling_rate: 16000
      - name: timestamps_start
        sequence: float64
      - name: timestamps_end
        sequence: float64
      - name: speakers
        sequence: string
    splits:
      - name: data
        num_bytes: 2124232888
        num_examples: 120
    download_size: 2089622461
    dataset_size: 2124232888
  - config_name: eng
    features:
      - name: audio
        dtype:
          audio:
            sampling_rate: 16000
      - name: timestamps_start
        sequence: float64
      - name: timestamps_end
        sequence: float64
      - name: speakers
        sequence: string
    splits:
      - name: data
        num_bytes: 2338819328
        num_examples: 140
    download_size: 2298491715
    dataset_size: 2338819328
  - config_name: jpn
    features:
      - name: audio
        dtype:
          audio:
            sampling_rate: 16000
      - name: timestamps_start
        sequence: float64
      - name: timestamps_end
        sequence: float64
      - name: speakers
        sequence: string
    splits:
      - name: data
        num_bytes: 2159798942
        num_examples: 120
    download_size: 2119318800
    dataset_size: 2159798942
  - config_name: spa
    features:
      - name: audio
        dtype:
          audio:
            sampling_rate: 16000
      - name: timestamps_start
        sequence: float64
      - name: timestamps_end
        sequence: float64
      - name: speakers
        sequence: string
    splits:
      - name: data
        num_bytes: 2456735672
        num_examples: 140
    download_size: 2424929812
    dataset_size: 2456735672
  - config_name: zho
    features:
      - name: audio
        dtype:
          audio:
            sampling_rate: 16000
      - name: timestamps_start
        sequence: float64
      - name: timestamps_end
        sequence: float64
      - name: speakers
        sequence: string
    splits:
      - name: data
        num_bytes: 2336018713
        num_examples: 140
    download_size: 2307711237
    dataset_size: 2336018713
configs:
  - config_name: deu
    data_files:
      - split: data
        path: deu/data-*
  - config_name: eng
    data_files:
      - split: data
        path: eng/data-*
  - config_name: jpn
    data_files:
      - split: data
        path: jpn/data-*
  - config_name: spa
    data_files:
      - split: data
        path: spa/data-*
  - config_name: zho
    data_files:
      - split: data
        path: zho/data-*
license: cc-by-nc-sa-4.0
language:
  - en
  - ja
  - zh
  - de
  - es
tags:
  - speaker-diarization
  - speaker-segmentation
  - voice-activity-detection
extra_gated_fields:
  Company: text
  Country: country

Dataset Card for the Callhome dataset for speaker diarization

The CALLHOME Corpus is a collection of unscripted telephone conversations between native speakers in Chinese, English, German, Japanese and Spanish.

This is a processed version of the original Callhome dataset from the TalkBank corpora taken from here. It contains subsets in Chinese, English, German, Japanese and Spanish:

Note: This dataset has been processed using diarizers. It makes the dataset compatible with the diarizers library to fine-tune pyannote segmentation models. In particular, we only keep the annotated parts of the callhome dataset. For more information on how the dataset has been processed, refer to diarizers.

Example Usage 

from datasets import load_dataset
ds = load_dataset("diarizers-community/callhome", "jpn")
print(ds)

gives:

DatasetDict({
    train: Dataset({
        features: ['audio', 'timestamps_start', 'timestamps_end', 'speakers'],
        num_rows: 120
    })
})

Dataset source and citation:

Chinese Corpus:

  • Participants: 140
  • Type of Study: phone call
  • Location: China
  • Media type: audio
  • DOI: doi:10.21415/T54022

English

  • Participants: 120
  • Type of Study: naturalistic
  • Location: USA
  • Media type: audio
  • DOI: doi:10.21415/T5KP54

German

  • Participants: 100
  • Type of Study: phone call
  • Location: United States
  • Media type: audio
  • DOI: doi:10.21415/T56P4B

Japanese

  • Participants: 120
  • Type of Study: phone call
  • Location: United States
  • Media type: audio
  • DOI: doi:10.21415/T5H59V

Spanish

  • Participants: 120
  • Type of Study: phone call
  • Location: United States
  • Media type: audio
  • DOI: doi:10.21415/T51K54

Contribution

Thanks to @kamil-akesbi and sanchit-gandhi for the contribution.