🧠 Text Summarizer using T5 & FastAPI

📌 Overview

This project is a Text Summarization Web Application built using a fine-tuned T5 Transformer model. It takes long dialogues or text as input and generates a concise summary.

---

🚀 Features

• 🔹 Text summarization using Transformer (T5)
• 🔹 Clean and responsive web UI
• 🔹 FastAPI backend for real-time inference
• 🔹 Preprocessing + tokenization pipeline
• 🔹 Beam search for better summary generation

---

🏗️ Tech Stack

• Python
• FastAPI
• HuggingFace Transformers
• PyTorch
• HTML, CSS, JavaScript

---

📂 Project Structure

Text Summarizer/
│
├── app.py                 # FastAPI backend
├── index.html             # Frontend UI
├── saved_summary_model/   # Trained T5 model
│   ├── config.json
│   ├── model.safetensors
│   ├── tokenizer.json
│   └── ...
└── README.md

---

⚙️ Installation & Setup

1️⃣ Clone the repository

git clone https://github.com/your-username/text-summarizer.git
cd text-summarizer

2️⃣ Install dependencies

pip install fastapi uvicorn transformers torch

3️⃣ Run the application

uvicorn app:app --reload

---

🌐 Usage

👉 Open in browser:

http://127.0.0.1:8000

👉 API Docs:

http://127.0.0.1:8000/docs

---

🧪 Example

Input:

A: Hi, how are you?
B: I'm fine. Let's meet tomorrow.

Output:

They plan to meet tomorrow.

---

🧠 Model Details

• Model: T5 (Text-to-Text Transfer Transformer)
• Fine-tuned on: SAMSum Dataset
• Task: Dialogue Summarization
• Generation Method: Beam Search (num_beams=4)

---

📈 Future Improvements

• 🔹 Add ROUGE score evaluation
• 🔹 Support PDF / file summarization
• 🔹 Deploy on cloud (Render / AWS)
• 🔹 Improve model accuracy with larger dataset
• 🔹 Add user authentication

---

🤝 Contributing

Pull requests are welcome! For major changes, please open an issue first.

---

📜 License

This project is open-source and available under the MIT License.

---

👩‍💻 Author

Ankita Ghavate

---

⭐ If you like this project

Give it a ⭐ on GitHub!