---
language:
- ps
license: mit
pretty_name: Pashto Synthetic Speech Dataset Parquet (3k)
size_categories:
- 1K<n<10K
tags:
- audio
- speech
- pashto
- tts
- synthetic
- parquet
task_categories:
- automatic-speech-recognition
metrics:
- wer
paperswithcode:
  eval_results: []
  preferred_citation: |-
    @dataset{pashto_synthetic_speech_parquet_3k,
      title = {Pashto Synthetic Speech Dataset Parquet (3k)},
      year = {2023}
    }
---

# Pashto Synthetic Speech Dataset Parquet (3k)

This dataset contains 6000 synthetic speech recordings in the Pashto language,
with 3000 male voice recordings and 3000 female voice recordings, stored in Parquet format.

## Dataset Information

- **Dataset Size**: 3000 sentences
- **Total Recordings**: 6000 audio files (3000 male + 3000 female)
- **Audio Format**: WAV, 44.1kHz, 16-bit PCM, embedded directly in Parquet files
- **Dataset Format**: Parquet with 500MB shards
- **Sampling Rate**: 44.1kHz (44100 Hz)
- **Voices**: 
  - Male: ps-AF-GulNawazNeural
  - Female: ps-AF-LatifaNeural
- **Total Audio Duration**: 
  - Male: 1596.27 seconds
  - Female: 1635.43 seconds
- **Average Clip Duration**:
  - Male: 0.53 seconds
  - Female: 0.55 seconds

## Dataset Structure

```
pashto_speech_parquet_3k/
  ├── data/                     # Parquet data files with embedded audio
  │   ├── shard-00000-of-....parquet
  │   ├── shard-00001-of-....parquet
  │   ├── ...
  │   ├── _schema.json          # Parquet schema definition
  │   └── _index.json           # Parquet metadata
  ├── dataset_info.json         # Dataset metadata
  └── README.md                 # This file
```

## Usage with HuggingFace Datasets

```python
from datasets import load_dataset

# Load the dataset
dataset = load_dataset("your-username/pashto_speech_parquet_3k")

# Access an example
example = dataset['train'][0]
print(example['text'])            # The transcript
print(example['audio'])           # Audio data
print(example['speaker'])         # Speaker gender
print(example['sampling_rate'])   # Audio sampling rate
```

## Usage Applications

This dataset can be used for:
- Training ASR (Automatic Speech Recognition) systems for Pashto
- Voice activity detection
- Speaker verification
- Speech synthesis evaluation

## License

This dataset is licensed under the MIT License.

## Citation

If you use this dataset in your research, please cite:

```
@dataset{pashto_synthetic_speech_parquet_3k,
  title = {Pashto Synthetic Speech Dataset Parquet (3k)},
  year = {2023},
  publisher = {HuggingFace},
  howpublished = {\url{https://huggingface.co/datasets/ihanif/pashto_speech_parquet_3k}}
}
```

## Acknowledgments

This dataset was created using Microsoft Edge TTS voices.