Sabtain-Dev/Pashto-ASR-Dataset

Pashto ASR Dataset

Domain-specific Pashto automatic speech recognition dataset.

Domains

• Agriculture
• General
• Food Services
• Health
• Services

Dataset Structure

• Kaggle: https://www.kaggle.com/datasets/itssabtain/pashto-asr-dataset

Domain/
├── audios/
└── transcripts.csv

Transcript Mapping

Audio files are mapped according to row order in the transcript CSV file.

Example:

• audio 1.wav → row 1 transcript
• audio 2.wav → row 2 transcript

Audio Formats

Multiple audio formats are included:

• mp3
• ogg
• m4a
• wav

The preprocessing pipeline converts all formats into WAV automatically.

Usage

Designed for:

• Whisper fine-tuning
• Pashto ASR research
• LoRA adaptation
• Domain-specific speech recognition

Author

Sabtain-Dev

---

license: mit