Text Summarization App

A web-based text summarization tool built with Flask and Hugging Face Transformers. Users can input articles and select from multiple fine-tuned models — T5 (CNN/DailyMail, PubMed, XSum), BART, and Pegasus — to generate concise summaries.

---

Features

• Summarize free-form articles
• Choose from multiple Transformer models:
  • T5
    • CNN/DailyMail
    • PubMed
    • XSum
  • BART (trained on all datasets combined)
  • Pegasus (trained on all datasets combined)
• Clean, responsive user interface
• Efficient backend using PyTorch and Transformers

---

Setup Instructions

1. Clone the repository

   git clone https://github.com/HassanSiddique2967/Abstractive-Text-Summarization-for-News-Research-Papers.git
   cd text-summarization-app

2. Create and activate a virtual environment

   python -m venv summarizer
   summarizer\Scripts\activate  # On Windows
   # OR
   source summarizer/bin/activate  # On macOS/Linux

3. Install dependencies

   pip install -r requirements.txt

4. Download or fine-tune models
   Place your fine-tuned models in the following directories:

   • t5_cnn_dailymail_finetuned/
   • t5_pubmed_finetuned/
   • t5_xsum_finetuned/
   • bart_finetuned_all/
   • pegasus_finetuned_all/

5. Run the app

   python app.py

   The app will start at: http://127.0.0.1:5000/

---

🧪 Model Details

Model	Dataset Used	Description
T5	CNN/DailyMail	News articles summarization
T5	PubMed	Scientific/biomedical summarization
T5	XSum	Extreme summarization with a single sentence
BART	All datasets	Combined training for general summarization
Pegasus	All datasets	Combined training for high-quality abstractive summarization

---

Requirements

• Python 3.8+
• Flask
• torch
• transformers

Refer to requirements.txt for the full list.

---

Acknowledgements

• Hugging Face Transformers
• Pretrained datasets: CNN/DailyMail, PubMed, XSum

---