metadata

dataset_info:
  features:
    - name: ID
      dtype: int64
    - name: kural
      dtype: string
    - name: audio
      dtype: audio
    - name: adhigaram
      dtype: string
    - name: paal
      dtype: string
  splits:
    - name: train
      num_bytes: 54954027.39
      num_examples: 1330
  download_size: 54573065
  dataset_size: 54954027.39
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: mit
language:
  - ta
tags:
  - tamil
  - thirukural
  - audio

Thirukural Voice-Based Dataset

📖 About the Dataset

This dataset contains voice recordings of all 1330 thirukural couplets, generated using the gTTS (Google Text-to-Speech) Python package. It aims to make thirukural more accessible for AI/ML applications, educational tools, and Tamil language preservation efforts.

🏛 Dataset Contents

Each record in this dataset includes:

ID – The unique number of the Kural (1 to 1330)
Kural – The original Tamil couplet
Audio – Pre-generated audio file (MP3 format)
Adhigaram – The chapter to which the Kural belongs
Paal – The broader section of thirukural (அறத்துப்பால், பொருட்பால், காமத்துப்பால்)

🎯 Why This Dataset?

🔹 Structured & Ready-to-Use

Unlike Google’s TTS, which generates speech on demand, this dataset provides pre-generated, structured data for direct use in research and applications.

🔹 Offline Accessibility

No internet dependency – the audio files are locally available for any project.

🔹 Consistency & Reliability

TTS models evolve over time, but this dataset ensures uniform pronunciation and consistent quality across applications.

🔹 Open & Community-Driven

An open-source resource for researchers, developers, and educators working on Tamil language AI, NLP, and speech synthesis.

🛠 Usage

Install Dependencies

To load this dataset in Python:

from datasets import load_dataset

dataset = load_dataset("Selvakumarduraipandian/thirukural-audio")

Access Data

Each record is structured as follows:

for data in dataset["train"]:
    print(f"ID: {data['ID']}")
    print(f"Kural: {data['kural']}")
    print(f"Audio File: {data['audio']}")
    print(f"Adhigaram: {data['adhigaram']}")
    print(f"Paal: {data['paal']}")

📢 Contributors

This dataset was created by

we used Python and gTTS.

📌 License

This dataset is open-source and free to use for research and educational purposes. If you use this dataset, please credit thirukural Voice-Based Dataset.

💬 Feedback & Contributions

We welcome feedback and contributions! Feel free to reach out if you'd like to improve or expand this dataset. 🚀

#thirukural #Tamil #AI #MachineLearning #SpeechSynthesis