README.md

---
task_categories:
- automatic-speech-recognition
language:
- ml
dataset_info:
  features:
  - name: audio_filepath
    dtype: audio
  - name: text
    dtype: string
  - name: duration
    dtype: float64
  - name: lang
    dtype: string
  splits:
  - name: train
    num_bytes: 19457087381.903
    num_examples: 224323
  download_size: 19431236298
  dataset_size: 19457087381.903
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: cc-by-4.0
pretty_name: Shruthilipi Malayalam Subset
---
# Shrutilipi-ML Dataset

## Overview
This dataset is a specific subset of the **Shrutilipi** ASR (Automatic Speech Recognition) corpus, containing only the **Malayalam** language data. 

Shrutilipi is a large-scale multilingual speech dataset for Indian languages, originally curated by AI4Bharat. This repository aims to provide a lightweight, language-specific version for researchers and developers focusing on Malayalam speech technology.

## Dataset Details
- **Source Dataset:** [ai4bharat/Shrutilipi](https://huggingface.co/datasets/ai4bharat/Shrutilipi)
- **Language:** Malayalam (ml)
- **Subset size:** ~16.4 GB
- **Task:** Automatic Speech Recognition (ASR)

## Usage
You can load this dataset directly using the Hugging Face `datasets` library:

```python
from datasets import load_dataset

dataset = load_dataset("trysem/Shrutilipi-ML")
print(dataset)
```
## Dataset Structure
The dataset typically contains:
- `audio`: The audio file/array.
- `sentence`: The corresponding Malayalam transcription.
- `id`: Unique identifier for the sample.

## Citation and Credits
If you use this dataset, please cite the original AI4Bharat Shrutilipi paper:

```bibtex
@article{bharat2024shrutilipi,
  title={Shrutilipi: A Massive Multilingual Corpus for Indian ASR},
  author={AI4Bharat},
  journal={arXiv preprint},
  year={2024}
}
```

## License
This dataset follows the licensing of the original Shrutilipi dataset (Creative Commons Attribution 4.0 International - CC BY 4.0).