README.md

metadata

dataset_info:
  features:
    - name: audio
      dtype:
        audio:
          sampling_rate: 22050
    - name: text
      dtype: string
    - name: speaker_id
      dtype: string
    - name: duration
      dtype: float64
  splits:
    - name: test_cs
      num_bytes: 72241660
      num_examples: 300
    - name: train_cs
      num_bytes: 1312289473
      num_examples: 5212
    - name: validation_cs
      num_bytes: 41637605
      num_examples: 165
    - name: test_to
      num_bytes: 60639242
      num_examples: 344
    - name: train_to
      num_bytes: 417689267
      num_examples: 2128
    - name: validation_to
      num_bytes: 12545256
      num_examples: 67
    - name: test_wavs
      num_bytes: 132818630
      num_examples: 644
  download_size: 1985213237
  dataset_size: 2049861133
configs:
  - config_name: default
    data_files:
      - split: test_cs
        path: data/test_cs-*
      - split: train_cs
        path: data/train_cs-*
      - split: validation_cs
        path: data/validation_cs-*
      - split: test_to
        path: data/test_to-*
      - split: train_to
        path: data/train_to-*
      - split: validation_to
        path: data/validation_to-*
      - split: test_wavs
        path: data/test_wavs-*
license: cc-by-nc-4.0
task_categories:
  - text-to-speech
language:
  - ar
tags:
  - code-switching
  - multilingual-tts
pretty_name: 'TunSwitch: A comprehensive Tunisian Arabic code-switched text-to-speech datase'
size_categories:
  - 10K<n<100K

---

Dataset: TunSwitch TTS
Language: Tunisian Arabic (Code-Switched)
Size: 8,860 audio samples (12.9 hours)
Sampling Rate: 22,050 Hz
Format: PCM 16-bit WAV
Splits: - TunSwitchCS: 5,677 samples (train/dev/test) - TunSwitchTO: 2,539 samples (train/dev/test)
- Test WAVs: 644 samples Features: Audio, text transcriptions, speaker IDs, durations License: CC-BY-NC-4.0

@misc{abdallah2023leveraging, title={Leveraging Data Collection and Unsupervised Learning for Code-switched Tunisian Arabic Automatic Speech Recognition}, author={Ahmed Amine Ben Abdallah and Ata Kabboudi and Amir Kanoun and Salah Zaiem}, year={2023}, eprint={2309.11327}, archivePrefix={arXiv}, primaryClass={eess.AS} }

@dataset{tunswitch_tts_2024, title = {TunSwitch TTS: Tunisian Arabic Code-Switched Speech Dataset}, author = Abdelkader Batnini, year = {2024}, publisher = {Abdelkader Batnini}, url = {https://huggingface.co/datasets/Batnini/tunswitch-tts}, version = {1.0}, license = {CC-BY-NC-4.0}, description = {A comprehensive Tunisian Arabic code-switched text-to-speech dataset containing 8,860 audio samples (12.9 hours) with transcriptions, speaker IDs, and duration information.} }