Welcome to Lexicon Generator

This project is a grapheme-to-phoneme (G2P) converter for Urdu language. It can generate lexicons for Urdu words using a deep learning model.

Files

• main.py: This is a summarized inference code implementation.
• app.py: This is a complete flask app and an API to generate lexicons.

Web App

You can run this app by the following command:

python app.py

This will open a nice web UI for you.

API Endpoint

The API is also defined in app.py and when we run the web UI, the API also runs. You can access the API at http://localhost:5000/g2p using a POST request. This API requires a JSON object like this:

{"text": "رکوا\nایران\nخرید"}

Each word is separated by a newline character (\n).

Steps to Run this Model

1. Create an Anaconda environment with Python 3.6.

conda create -n env_name python=3.6 anaconda

2. Reach inside this directory and activate the conda environment:

conda activate env_name

3. Run the following command to install TensorFlow 1.0.0:

pip install tensorflow-1.0.0-cp36-cp36m-win_amd64.whl

4. Run the following command to install other requirements:

pip install -r requirements.txt

5. Now you are ready to go:

python app.py

I hope it was helpful. 😊