Skip to content

Latest commit

 

History

History
61 lines (41 loc) · 3.86 KB

README.md

File metadata and controls

61 lines (41 loc) · 3.86 KB

Language Translator Application

This project is a Flask-based web application that translates text from one language to another using a pre-trained machine translation model (NLLB-200). The application provides a user-friendly interface for selecting languages and entering text, and it returns the translated output along with the computation time.

Flask App

Overview

The application allows users to translate text by:

  1. Selecting the input language. (150+ language support)
  2. Selecting the target output language. (150+ language support)
  3. Entering the text they wish to translate.
  4. Submitting the request to view the translated text and computation time.

Features

  • Translation Pipeline: The app leverages the NLLB-200 (No Language Left Behind) model for accurate translation across 200 languages. This model, developed by Meta, is specifically optimized for low-resource languages, making it ideal for global use.
  • Language Selection: The app dynamically fetches supported languages from the backend, allowing users to select from the available options.
  • Spinner for Loading Indicator: To improve user experience, a spinner appears when the translation process is ongoing, and it disappears once the translation is complete.
  • Computation Time: After translation, the time taken to perform the translation is displayed.

Technology Stack

  • Flask: Used as the backend framework to handle HTTP requests, render templates, and communicate with the translation model.
  • HTML, CSS, and JavaScript: The frontend is built with HTML and CSS for styling and JavaScript (AJAX) for handling asynchronous requests.
  • Hugging Face Transformers: The NLLB-200 model is loaded via the Hugging Face Transformers library, using a pre-trained model for multi-lingual text translation. Visit HF model link for downloading the model
  • CUDA Support: If available, the application utilizes GPU processing via CUDA to speed up the translation process.

Model: NLLB-200

The NLLB-200 model, "No Language Left Behind," is a state-of-the-art machine translation model developed by Meta, supporting translations across 200 languages, especially for low-resource languages. By using a sequence-to-sequence model architecture, NLLB-200 achieves high translation quality by leveraging a massive multilingual dataset and advanced training techniques.

This model provides:

  • High accuracy across multiple language pairs.
  • Low latency, making it well-suited for applications requiring quick responses.
  • Language codes for each supported language, allowing flexible input and output language options.

File Structure

  • flask_app.py: Main application file handling routing, translation processing, and language list retrieval.
  • app.py: it is the api code for deployment purpose
  • templates/index.html: Frontend layout for user interaction with the translation app.
  • static/style.css: Custom CSS for styling the frontend interface.
  • nllb_supported_languages.py: Contains language data, used for language selection in the frontend.

Setup Instructions

  1. Install Python Packages: Make sure to install Flask, Transformers, and Torch (for GPU support if available).

    pip install flask transformers torch

  2. Download and Save Model: Download the NLLB-200 model and save it in your desired directory. This application is set to load the model locally from D:/Language Translator/model you can save it anywhere you want just pass the path in the *pretrained_model_name_or_path* parameter.

  3. Run the Application:

    python flask_app.py

For any queries you can contact me through my mail id mail or text me in linkedin LinkedIn