README.md

# MLS-Sidon

## Overview
This dataset is a **cleansed version of Multilingual LibriSpeech (MLS)** with **Sidon** speech restoration mode for **Speech Synthesis** and **Spoken Language Modeling**.  

The dataset is provided in **[WebDataset](https://github.com/webdataset/webdataset) format** for efficient large-scale training.  

- **Source**: [Multilingual LibriSpeech](https://www.openslr.org/94/)
- **Languages**: English, German, French, Spanish, Italian, Portuguese, Polish, Dutch  
- **Format**: WebDataset (`.tar` shards)  
- **License**: [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)

---

## Dataset Structure
Each sample in the dataset contains:  

- **`flac`** — audio file (48 kHz, single channel)  
- **`meta.json`** *(optional)* — metadata including language, speaker ID, and original MLS reference  

Example (inside a `.tar` shard):  
```

000001.flac
000001.meta.json
000002.flac
000002.meta.json
...

````

---

## How to Use

### With 🤗 Datasets
You can load the WebDataset directly with Hugging Face’s `datasets` library:

```python
from datasets import load_dataset

ds = load_dataset(
    "webdataset",
    data_files="https://huggingface.co/datasets/<username>/<repo>/resolve/main/{lang}_shard-{000000..000099}.tar",
    split="train",
    streaming=True
)

for sample in ds:
    audio = sample["flac"]
    text = sample["txt"]
    print(audio, text)
````

Replace `{lang}` with the language (e.g., `english`, `german`).

---

### With WebDataset (PyTorch)

```python
import webdataset as wds

urls = "https://huggingface.co/datasets/<username>/<repo>/resolve/main/english_shard-{000000..000099}.tar"

dataset = (
    wds.WebDataset(urls)
    .decode()
    .to_tuple("flac", "txt")
)

for audio, text in dataset:
    ...
```

---

## Citation

If you use this dataset, please cite Sidon and the original MLS paper:

```
@misc{nakata2025sidonfastrobustopensource,
      title={Sidon: Fast and Robust Open-Source Multilingual Speech Restoration for Large-scale Dataset Cleansing}, 
      author={Wataru Nakata and Yuki Saito and Yota Ueda and Hiroshi Saruwatari},
      year={2025},
      eprint={2509.17052},
      archivePrefix={arXiv},
      primaryClass={cs.SD},
      url={https://arxiv.org/abs/2509.17052}, 
}
```


```
@inproceedings{pratap2020mls,
  title     = {MLS: A Large-Scale Multilingual Dataset for Speech Research},
  author    = {Pratap, Vineel and Xu, Qiantong and Sriram, Anuroop and others},
  booktitle = {Interspeech},
  year      = {2020}
}
```


---

## License

This dataset is released under [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/).

---

## Acknowledgements

* **Original data**: [Multilingual LibriSpeech (MLS)](https://www.openslr.org/94/)