---
dataset_info:
  features:
  - name: id
    dtype: int64
  - name: english_audio
    dtype:
      audio:
        sampling_rate: 16000
  - name: english_text
    dtype: string
  - name: english_duration
    dtype: float32
  - name: french_audio
    dtype:
      audio:
        sampling_rate: 16000
  - name: french_text
    dtype: string
  - name: french_duration
    dtype: float32
  - name: overlap_ratio
    dtype: float32
  splits:
  - name: train
    num_bytes: 11394031140.74
    num_examples: 10167
  download_size: 9821132950
  dataset_size: 11394031140.74
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: cc-by-nc-4.0
task_categories:
- text-to-speech
- automatic-speech-recognition
- audio-to-audio
language:
- en
- fr
---


# DuBLaB-en-fr-0.7-f-0.5 

###  Description

**DuBLaB-en-fr-0.7-f-0.5** is a bilingual dataset of **English-French audio segments**, aligned and extracted from audiovisual sources. This dataset is designed for **automatic speech recognition (ASR), machine translation, cross-lingual style mapping, and text-to-speech (TTS) applications**.

This version was built on top of the  **0.7 version**  and  **filtered**  to remove segments with a **word-per-second (WPS) density below 0.5**. The filtering process helps eliminate segments containing **excessive silence or background noise**, improving the dataset’s overall quality.


- **Repository:** [Hugging Face - DuBLaB-en-fr-0.7-f-0.5](https://huggingface.co/datasets/amine-khelif/DuBLaB-en-fr-0.7-f-0.5)

---

## Structure

Each dataset entry contains:

- **`id`**: Unique identifier for the segment.
- **`english_audio`**: Audio file (16 kHz sampling rate).
- **`english_text`**: English transcript.
- **`english_duration`**: Duration of the English segment.
- **`french_audio`**: Audio file (16 kHz sampling rate).
- **`french_text`**: French transcript.
- **`french_duration`**: Duration of the French segment.
- **`overlap_ratio`**: IOU overlap between the English and French segments.

---


### **Duration Statistics**

- English durations:
```
  Global Total Duration: 31409.2120 s
  Mean:                  3.0893 s
  Minimum:               0.2810 s
  Maximum:               29.9230 s
  Variance:              6.3013 s^2
```
- French durations:
```
  Global Total Duration: 33179.8130 s
  Mean:                  3.2635 s
  Minimum:               0.3430 s
  Maximum:               29.9900 s
  Variance:              6.8527 s^2
```
---

## Visualizations

### **Text Density Distribution**
This plot illustrates the **distribution of words-per-second density** after filtering.  

![f05.png](https://cdn-uploads.huggingface.co/production/uploads/65abc3bd302258937183a7d3/SYvvThsZ7E0qE84_xI83F.png)

---

## Usage

### **Loading the Dataset**
Use the Hugging Face `datasets` library:

```python
from datasets import load_dataset

dataset = load_dataset("amine-khelif/DuBLaB-en-fr-0.7-f-0.5")
print(dataset)
```
License

This dataset is released under the CC BY-NC 4.0 license.
It is free to use exclusively for non-commercial purposes.

## Citation

If you use **DuBLaB-en-fr-0.7-f-0.5** in your research, please cite it as follows:

```
@misc{DuBLaB-en-fr-0.7-f-0.5,
  author       = {Khelif, Amine Khelif},
  title        = {DuBLaB en-fr Dataset (Version 0.7-f-0.5)},
  howpublished = {\url{https://huggingface.co/datasets/amine-khelif/DuBLaB-en-fr-0.7-f-0.5}},
  year         = {2025},
  note         = {Released under the CC BY-NC 4.0 license}
}
```




## Contact

For any questions, suggestions, or contributions, please contact:
- Author: Amine Khelif KHELIF
- Email: khelifaminekhelif@gmail.com


🚀 Start exploring DuBLaB-en-fr-0.7-f-0.5 today! 🚀