Aananda-giri/openSLR-Nepali

OpenSLR Nepali Speech Dataset (Preprocessed)

Dataset Description

This is a preprocessed version of the Nepali speech dataset from OpenSLR, ready for training speech models including Automatic Speech Recognition (ASR) and Text-to-Speech (TTS).

Dataset Statistics

• Total Audio Files: 118,231
• Total Duration: 57.34 hours
• Sample Rate: 16kHz
• Channels: Mono
• Format: WAV

Preprocessing Applied

1. Text Preprocessing:

   • Text cleaning and normalization
   • Number to text conversion
   • Nepali character standardization

2. Audio Preprocessing:

   • Resampled to 16kHz mono
   • Quality filtering (duration: 0.5-20s, SNR > 10dB)
   • Silence detection and handling
   • RMS normalization
   • Pre-emphasis filter (coefficient=0.97)

3. Quality Checks:

   • Verified all audio files exist
   • Removed empty transcripts
   • Duration validation

Dataset Structure

.
├── metadata.csv          # Contains file_name, transcript, duration, sample_rate, gender, quality, snr_db
└── audio/               # Directory containing all audio files
    └── *.wav            # Audio files

metadata.csv Format

Column	Description
file_name	Relative path to audio file (audio/filename.wav)
transcript	Nepali text transcript
duration	Audio duration in seconds
sample_rate	Sample rate (16000 Hz)
gender	Speaker gender (male/female)
quality	Audio quality rating (high/low)
snr_db	Signal-to-noise ratio in decibels

Usage

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("Aananda-giri/openSLR-Nepali")

# Access a sample
sample = dataset['train'][0]
print(f"Audio: {sample['audio']}")
print(f"Transcript: {sample['transcript']}")
print(f"Duration: {sample['duration']} seconds")

Use Cases

This dataset is suitable for:

• Automatic Speech Recognition (ASR): Train models to transcribe Nepali speech to text
• Text-to-Speech (TTS): Train models to synthesize Nepali speech from text
• Speech Research: Study Nepali phonetics, prosody, and acoustic characteristics

Source

Original dataset from OpenSLR:

• OpenSLR Nepali Speech Dataset

License

This dataset is released under the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0).

Citation

If you use this dataset, please cite the original OpenSLR dataset:

@misc{openslr_nepali,
  title={Nepali Speech Dataset},
  author={OpenSLR},
  howpublished={\url{https://www.openslr.org/}},
  year={2020}
}

Preprocessing Details

This dataset was preprocessed using a custom pipeline that includes:

• Text normalization for Nepali language
• Audio quality filtering and enhancement
• Standardization to 16kHz mono format
• Vocabulary and character mapping generation

Preprocessed on: 2025-11-17

Contact

For questions or issues with this preprocessed version, please open an issue in the dataset repository.