Nepali ASR Dataset

Overview

This dataset is created to support Automatic Speech Recognition (ASR) tasks for the Nepali language. It consists of Nepali speech data collected from various open-source platforms, such as Mozilla OpenVoice Dataset, OpenSLR, and other publicly available resources. The primary goal of this dataset is to enhance ASR model performance for Nepali, a language with limited representation in existing ASR datasets.

Key Features

Language: Nepali
Data Type: Speech-to-text pairs
Format: Audio files and corresponding transcriptions
Sources: Mozilla OpenVoice Dataset, OpenSLR, and other open platforms
Use Case: Training and evaluation of ASR models for Nepali
Average Audio Length: 15-30 seconds

Structure

The dataset is organized as follows:

nepali-speech-to-text/
|-- audio/
|   |-- file1.wav
|   |-- file2.wav
|   |-- file3.wav
|-- audio_transcripts.csv

audio/: Contains all the speech audio files in .wav format.
audio_transcripts.csv: A CSV file containing the transcriptions for the audio files.

Licensing and Permissions

The dataset is compiled from publicly available open-source data and follows the respective licenses of its sources. Ensure compliance with each source's licensing terms when using this dataset.

Usage

This dataset can be used for:

Training ASR models for Nepali.
Fine-tuning existing speech-to-text models like Whisper.
Benchmarking ASR systems for Nepali.

Contributions

Contributions to enhance the dataset are welcome! You can:

Add more Nepali speech data.
Provide better transcriptions.
Report any issues or errors in the existing dataset.

Citation

If you use this dataset in your research or projects, please cite it appropriately. Include attribution to the original sources as well.

Mozilla Common Voice

Website: https://commonvoice.mozilla.org/en/datasets

OpenSLR

@inproceedings{kjartansson-etal-sltu2018,
    title = {{Crowd-Sourced Speech Corpora for Javanese, Sundanese,  Sinhala, Nepali, and Bangladeshi Bengali}},
    author = {Oddur Kjartansson and Supheakmungkol Sarin and Knot Pipatsrisawat and Martin Jansche and Linne Ha},
    booktitle = {Proc. The 6th Intl. Workshop on Spoken Language Technologies for Under-Resourced Languages (SLTU)},
    year  = {2018},
    address = {Gurugram, India},
    month = aug,
    pages = {52--55},
    URL   = {http://dx.doi.org/10.21437/SLTU.2018-11}
}

Contact

For questions, suggestions, or contributions, please contact [[email protected]].

Disclaimer: While the dataset is compiled with care, it may still contain errors or inconsistencies. Users are encouraged to verify the data before use.

audio audioduration (s) 1.31 30