Text Summarizer App

A simple web application built with FastAPI and Hugging Face Transformers that uses the T5 model to generate concise summaries from dialogue or long text.

Features

• Data Training Pipeline: Includes a Jupyter Notebook for data preprocessing and model training.
• T5-based Summarization: Uses a fine-tuned T5 model for high-quality text compression.
• FastAPI Backend: High-performance asynchronous API.
• Jinja2 Templates: Simple HTML interface for easy interaction.
• Hardware Acceleration: Automatically detects and uses CUDA (NVIDIA) or MPS (Apple Silicon) if available, falling back to CPU.

Workflow

graph TD
    A[CSV Data] --> B[Notebook: Preprocessing]
    B --> C[Notebook: Model Training]
    C --> D[Saved Model Artifacts]
    D --> E[FastAPI /summarize Endpoint]
    E --> F[Data Cleaning / Regex]
    F --> G[T5 Tokenization]
    G --> H[Model Inference]
    H --> I[Return Summary JSON]

Loading

Project Structure

.
├── app.py                  # FastAPI application logic
├── index.html              # Frontend UI
├── text_summarizer.ipynb   # Data preprocessing & training pipeline
├── data/                   # Source .csv datasets for training
├── saved_summary_model/    # Local T5 model weights and tokenizer (output of notebook)
└── venv/                   # Python virtual environment

Setup & Installation

1. Clone the repository:

   git clone github.com/imHardik1606/T5-Text-Summarizer
   cd "Summarizer App"

2. Create and activate a virtual environment:

   python -m venv venv
   .\venv\Scripts\Activate.ps1

3. Install dependencies:

   pip install fastapi uvicorn transformers torch jinja2 sentencepiece

4. Add your model: Place your fine-tuned T5 model files inside the saved_summary_model/ folder.

Running the App

Start the server using uvicorn:

uvicorn app:app --reload

Open your browser and navigate to http://127.0.0.1:8000.

API Endpoints

• GET /: Serves the web-based UI.
• POST /summarize: Accepts a JSON body {"dialogue": "your text here"} and returns a summary.

License

MIT