masakachristopher / sukuma-translator-api

Translation API with Flask

Overview

This is a Flask-based API that translates input text using a pre-trained TensorFlow model. The API accepts POST requests with input text and returns the translated output.

Translates english and swahili phrases to sukuma tribe language

Installation

1. Clone the repository:

   git clone https://github.com/masakachristopher/translator-flask-server.git

2. Change directory to the root folder of your project:

   cd translator-flask-server

3. Create virtual environment

   python3 -m venv venv
   source venv/bin/activate

4. Install dependencies

   pip install -r requirements.txt

5. Download the resorces here sukuma-translation and tokens folders

   Download Resources

6. Create folder with name models in the project root folder and place the sukuma-translation folder inside it. If named otherwise, such as trans please rename it to sukuma-translation

   - Root
       - models
           - sukuma-translation
       - src
       - app
       // others root files or folders
   

7. Place folder with name tokens in the project root folder which contains the tokenizer files inside it

   - Root
       - tokens
           - input_tokenizer.pkl
           - target_tokenizer.pkl
       - src
       - app
       // others root files or folders
   

Usage

1. Run your app

   flask run

2. Make request

   • URL: http://localhost:5000/translate

   • Headers: Content-Type: application/json

   • Body (Json Payload)

     {
     "text": "water"
     }

     or

     {
     "text": "hapana"
     }

   • Response sample

     {
     "translated_text": "minze"
     }

3. Customize the model and translations:

   • If you want to use a different pre-trained TensorFlow model, replace the trans file with your own trained model. Update the file path in the Flask app accordingly.
   • If you want to modify the translations, update the target_texts_combined in the Flask app with your desired translations

Contributing

Contributions are welcome! If you find any issues or have suggestions for improvements, please open an issue or submit a pull request.