English to Spanish Translator with Streamlit

This project demonstrates how to build an interactive English-to-Spanish translation app using a fine-tuned MarianMT model and the Streamlit framework.

Table of Contents

• Description
• Features
• Installation
• Docker
• Usage
• Demo Video
• Model Details
• Evaluation
• Future Work
• License

Description

This Streamlit app provides a simple yet effective interface for translating English text into Spanish. It leverages a state-of-the-art MarianMT model fine-tuned on a large parallel corpus of English-Spanish text. The app features a progress bar to indicate translation progress and handles potential errors gracefully.

Features

• Real-time Translation: Translate English text to Spanish instantly.
• Progress Bar: Visual feedback during translation.
• Error Handling: Gracefully handles errors during translation.
• User-friendly Interface: Easy-to-use Streamlit interface.

Installation

1. Clone the Repository:

   git clone [https://github.com/WizKnight/English-to-Spanish-Translation-APP.git](https://github.com/WizKnight/English-to-Spanish-Translation-APP.git)
   

2. Install Dependencies:

   pip install -r requirements.txt
   

3. Run Jupyter Notebook:
   • Train the model using Torch cuda and save the model.

Dockers

1. Build Docker Image:

   docker build -t english-to-spanish-translator . 
   

2. Run Docker Container:

   docker run -p 8501:8501 english-to-spanish-translator
   

Usage

1. Run the App:

   streamlit run main.py
   

2. Enter Text: Type or paste your English text into the text area.
3. Click Translate: Press the "Translate" button to initiate translation.
4. View Translation: The translated Spanish text will appear below.

Demo Video

Model Details

• Architecture: MarianMT (Helsinki-NLP/opus-mt-en-es)
• Fine-tuning: The model has been fine-tuned on a large parallel corpus of English-Spanish text from the Hugging Face Datasets library.
• Framework: PyTorch and Transformers (Hugging Face)

Evaluation

• The model was evaluated using the BLEU (Bilingual Evaluation Understudy) score, achieving a score of 22.96 on a test dataset.
• Additional human evaluation was conducted to assess translation fluency and adequacy.

Future Work

• Add support for batch translation.
• Integrate a download option for translated text.
• Explore other model architectures like T5.
• Incorporate user feedback for continuous improvement.

Liscense

This project is licensed under the Apache 2.0 License.