Text Summarizer App

A web application that summarizes dialogue and text content using a fine-tuned T5 transformer model, served through a FastAPI backend.

Overview

This project provides a simple interface for pasting text or dialogue and receiving a concise summary in return. The summarization is powered by a T5 (Text-to-Text Transfer Transformer) model, and the application is built with a FastAPI backend and a lightweight HTML/CSS/JavaScript frontend.

Features

• Text summarization using a fine-tuned T5 model
• Simple, responsive web interface
• REST API endpoint for programmatic access
• Input cleaning and preprocessing before summarization
• Beam search decoding for improved summary quality

Tech Stack

Backend

• Python
• FastAPI
• Uvicorn (ASGI server)
• Hugging Face Transformers
• PyTorch

Frontend

• HTML
• CSS
• Vanilla JavaScript

Project Structure

Text-Summarizer-App/
├── app.py                          # FastAPI application and model logic
├── index.html                      # Frontend page
├── static/
│   ├── script.js                   # Frontend logic (form handling, API calls)
│   └── style.css                   # Styling
├── text-summarizer_saved_summary_model/
│   ├── config.json
│   ├── generation_config.json
│   ├── model.safetensors
│   ├── tokenizer_config.json
│   └── tokenizer.json
├── requirements.txt                # Python dependencies
└── README.md

Getting Started

Prerequisites

• Python 3.11 or later
• pip

Installation

1. Clone the repository

   git clone https://github.com/Samir-BK/Text-Summarizer-App.git
   cd Text-Summarizer-App

2. Install dependencies

   pip install -r requirements.txt

3. Run the application

   python -m uvicorn app:app --reload

4. Open your browser and navigate to

   http://127.0.0.1:8000
   

API Usage

POST /summarize/

Summarizes the provided text.

Request body

{
  "dialogue": "Your text or dialogue to summarize goes here."
}

Response

{
  "summary": "Generated summary text."
}

How It Works

1. Input text is cleaned (whitespace normalization, HTML tag removal, lowercasing).
2. The cleaned text is tokenized and passed to the T5 model.
3. The model generates a summary using beam search decoding.
4. The generated token sequence is decoded back into readable text and returned to the frontend.

Notes

• The model files are loaded locally from the text-summarizer_saved_summary_model/ directory.
• torch will automatically use GPU (CUDA or MPS) if available, otherwise it falls back to CPU.

License

This project is provided as-is for educational and demonstration purposes.