README.md

metadata

language:
  - bo
  - en
multilinguality:
  - translation
size_categories:
  - n<1K
task_categories:
  - automatic-speech-recognition
  - translation
  - text-to-speech
pretty_name: Tibetan Audio-English Sentence Dataset (Sample)
tags:
  - tibetan
  - audio
  - speech
  - translation
  - sample
dataset_info:
  features:
    - name: audio
      dtype:
        audio:
          sampling_rate: 16000
    - name: sentence
      dtype: string
  splits:
    - name: train
      num_bytes: 1542386
      num_examples: 5
  download_size: 1494412
  dataset_size: 1542386
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Tibetan Audio-English Sentence Dataset (Sample)

This is a sample dataset containing 5 rows from a merged collection of 6 Tibetan audio datasets with English translations.

📊 Dataset Details

• Total Samples in Full Dataset: 17,278
• Samples in This Preview: 5
• Format: Audio + English sentence pairs
• Audio Sampling Rate: 16,000 Hz
• Languages: Tibetan (audio) → English (text)

🗂️ Source Datasets

This sample is merged from 6 datasets:

1. lilgoose777/tibetan-speech-english-text-dataset (3,921 samples)
2. lilgoose777/merged-tibetan-titung-goose (5,188 samples)
3. Titung/tibetan-to-english-audio-dataset (1,178 samples)
4. Dataset 4 (294 samples)
5. Dataset 5 (294 samples)
6. Dataset 6 (6,403 samples)

📋 Dataset Structure

{
    'audio': Audio(sampling_rate=16000),  # Tibetan speech audio
    'sentence': string                     # English translation
}

🚀 Usage

from datasets import load_dataset

# Load the sample dataset
dataset = load_dataset("lilgoose777/tibetan-audio-english-6datasets-sample")

# Access first sample
sample = dataset['train'][0]
audio = sample['audio']
sentence = sample['sentence']

print(f"Sentence: {sentence}")
print(f"Audio array shape: {audio['array'].shape}")
print(f"Sampling rate: {audio['sampling_rate']}")

💡 Purpose

This is a preview/test dataset to verify:

• Dataset structure and format
• Audio loading and decoding
• Data quality

For the full dataset with all 17,278 samples, please contact the dataset owner.

📝 License

Please refer to the original source datasets for licensing information.

🙏 Acknowledgments

Special thanks to the creators of the original datasets used in this merge.