metadata

dataset_info:
  features:
    - name: audio_name
      sequence: string
    - name: dialect
      sequence: string
    - name: gender
      sequence: string
    - name: speaker
      sequence: string
  splits:
    - name: train
      num_bytes: 296025
      num_examples: 6941
    - name: validation
      num_bytes: 52245
      num_examples: 1225
    - name: clean_test
      num_bytes: 64000
      num_examples: 1500
    - name: noise_test
      num_bytes: 43520
      num_examples: 1020
  download_size: 87783
  dataset_size: 455790
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: clean_test
        path: data/clean_test-*
      - split: noise_test
        path: data/noise_test-*
language:
  - vi
tags:
  - audio
  - voice-classification
  - dialect-gender-recognition

Dataset Card for the ViSpeech dataset

Dataset Summary

The ViSpeech Dataset is a collection of unscripted audio recordings designed for the classification of gender and Vietnamese dialects. The dataset comprises 10,686 mp3 files, totaling slightly over 14 hours of speech data from 449 speakers representing both genders across the three primary Vietnamese dialects: Northern, Central, and Southern. It is divided into three subsets: a training set with clean recordings and two test sets—one with clean recordings and the other with ambient noise. Notably, the speakers in the training set are independent of those in the test sets. The dataset is designed to provide a diverse and comprehensive resource for audio classification research.

Dataset Details

Dataset Information

Configurations: The dataset contains the following configurations:
- train: Training data with gender and dialect information.
- validation: Validation data with gender and dialect information.
- clean_test: Clean test data for evaluation.
- noise_test: Noisy test data for evaluation.
Features:
- audio: Audio data with a sampling rate of 16,000 Hz.
- audio_path: File path to each audio file.
- audio_name: The name of the audio file.
- dialect: The dialect of the speaker.
- gender: The gender of the speaker.
- speaker: The speaker's identifier.
Splits:
- Train: Contains the training data files.
- Validation: Contains the validation data files.
- Clean Test: Contains the clean test data files.
- Noise Test: Contains the noisy test data files.

Licensing

The dataset is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 (cc-by-nc-sa-4.0).

Example Usage

You can load the dataset in Python using the datasets library:

from datasets import load_dataset
ds = load_dataset("{dataset_name}", "train")
print(ds)

gives:

DatasetDict({
    train: Dataset({
        features: ['audio_name', 'dialect', 'gender', 'speakers'],
        num_rows: 6941
    })
})