This project is designed to collect and manage voice clips from the public to create a comprehensive Nepali Roman text dataset. The dataset is intended for use in training speech recognition models to understand and transcribe Nepali spoken in Roman script.
Live URL: https://speech.subasacharya.com.np/
- Record: Users can record and upload audio clips directly through the web interface.
- Verify: Users can verify the accuracy of transcriptions to ensure the dataset's quality.
- Write: Users can submit new sentences to be recorded.
- Clone the repository:
git clone https://github.com/Subu19/Nepali-Roman-Dataset-Generator.git
- Navigate to the project directory:
cd Nepali-Roman-Dataset-Generator
- Install the dependencies:
npm install
- Start the server:
node index.js
Create a config.json
file in the project root with the following content:
{
"uri": "Your mongoDB URI",
"port": "port number"
}
Replace Your mongoDB URI
with your MongoDB connection string and port number
with the desired port number for the server.
- Record: Navigate to the recording page and use the provided tools to record and upload audio clips.
- Verify: Review and verify transcriptions to maintain the dataset's accuracy.
- Write: Submit new sentences to be recorded.
This project aims to build a robust dataset for improving Nepali speech recognition in the Roman script, enabling better language technology for Nepali speakers.