Text Summarizer

🧠 AI-Powered Dialogue Summarization using state-of-the-art transformer models.

Built by Parshv Patel | UC Berkeley • Data Science & ML

✨ Features

Backend

• Abstractive Summarization: Generates human-like summaries of conversations
• FLAN-T5 Model: Pre-trained on SAMSum with ROUGE-1 score of 47.24
• Production-Ready Pipeline: 5-stage ML pipeline (Ingestion → Validation → Transformation → Training → Evaluation)
• FastAPI REST API: Production-grade API with Swagger documentation
• Confidence Scoring: Returns confidence scores for generated summaries
• Length Control: Short, medium, and long summary options

Frontend

• Modern React UI: Built with React 19, Vite, and Tailwind CSS
• Premium Animations: Lenis smooth scroll, GSAP text animations, Framer Motion
• Glassmorphism Design: Beautiful dark theme with glass effects
• Interactive Features: Copy to clipboard, processing time display, compression visualization
• Responsive Design: Works on desktop, tablet, and mobile
• Accessibility: Reduced motion support, keyboard navigation

Architecture

Text-Summarizer/
├── config/ # Configuration files
│ ├── config.yaml # Paths, model names
│ └── params.yaml # Training hyperparameters
├── src/text_summarizer/
│ ├── components/ # Core ML components
│ │ ├── data_ingestion.py
│ │ ├── data_validation.py
│ │ ├── data_transformation.py
│ │ ├── model_trainer.py
│ │ └── model_evaluation.py
│ ├── pipeline/ # Pipeline orchestration
│ │ ├── stage_01_data_ingestion.py
│ │ ├── stage_02_data_validation.py
│ │ ├── stage_03_data_transformation.py
│ │ ├── stage_04_model_trainer.py
│ │ ├── stage_05_model_evaluation.py
│ │ └── inference.py # Inference pipeline + CLI
│ ├── config/ # Configuration management
│ ├── entity/ # Data classes
│ └── utils/ # Utilities
├── tests/ # Unit tests
├── research/ # Jupyter notebooks
├── main.py # Training entry point
└── app.py # FastAPI application

Quick Start

Installation

# Clone the repository
git clone https://github.com/ParshvCrafts/Text-Summarizer.git
cd Text-Summarizer
# Create virtual environment
python -m venv venv
source venv/bin/activate # Linux/Mac
venv\Scripts\activate # Windows# Install dependencies
pip install -r requirements.txt

Training Profiles

Choose a profile based on your hardware and time constraints:

Profile	Data	Time	Best For
zero_training	0%	2 min	Laptops - Uses pre-trained model, no training!
quick_test	10%	10-15 min	Testing the pipeline works
laptop_friendly	50%	30-60 min	Laptops with limited resources
full_training	100%	2-4 hrs	Cloud/powerful GPU

Set your profile in config/config.yaml:

training_profile: zero_training # Recommended for most users!

Zero-Training Mode (Recommended)

The default configuration uses philschmid/flan-t5-base-samsum, which is already fine-tuned on SAMSum with ROUGE-1 score of 47.24. No training needed!

# Just run inference directly
python -m src.text_summarizer.pipeline.inference --interactive

Training (If Needed)

# Run smoke test first to verify everything works
python smoke_test.py
# Then run training
python main.py

This will:

1. Download the SAMSum dataset
2. Validate the data
3. Tokenize dialogues and summaries
4. Fine-tune the model with early stopping
5. Evaluate and save ROUGE metrics

Features:

• ✅ Automatic checkpoint resumption (never lose progress!)
• ✅ Training time estimation before starting
• ✅ Gradient checkpointing (saves memory)
• ✅ Early stopping (prevents overfitting)

Inference

CLI Usage

# Summarize text directly
python -m src.text_summarizer.pipeline.inference --text "John: Hi! Sarah: Hello, how are you?"# Summarize from file
python -m src.text_summarizer.pipeline.inference --file dialogue.txt
# Interactive mode
python -m src.text_summarizer.pipeline.inference --interactive

Python API

fromsrc.text_summarizer.pipeline.inferenceimportSummarizersummarizer=Summarizer()
summary=summarizer.summarize(""" John: Hey, are you coming to the party tonight? Sarah: I'm not sure, I have work to do. John: Come on, it'll be fun! Sarah: Okay, I'll try to come by 8.""")
print(summary)

API Server (Optional)

# Start the FastAPI server
uvicorn app:app --reload --host 0.0.0.0 --port 8000
# Test the API
curl -X POST "http://localhost:8000/summarize" \
-H "Content-Type: application/json" \
-d '{"text": "John: Hi! Sarah: Hello!"}'

Configuration

Model Selection

Edit config/config.yaml to change models:

data_transformation:
tokenizer_name: facebook/bart-large-cnn # Must match model_ckptmodel_trainer:
model_ckpt: facebook/bart-large-cnn

Supported models:

• facebook/bart-large-cnn (recommended)
• facebook/bart-large-xsum
• google/pegasus-cnn_dailymail
• philschmid/bart-large-cnn-samsum (pre-fine-tuned)

Training Parameters

Edit config/params.yaml:

TrainingArguments:
num_train_epochs: 5learning_rate: 2e-5per_device_train_batch_size: 2early_stopping_patience: 5

Testing

# Run all tests
pytest
# Run with coverage
pytest --cov=src

Project Structure

Component	Description
data_ingestion	Downloads and extracts SAMSum dataset
data_validation	Verifies train/test/validation splits exist
data_transformation	Tokenizes text using model's tokenizer
model_trainer	Fine-tunes with early stopping, LR scheduling
model_evaluation	Computes ROUGE, generates samples
inference	Production inference with caching

Performance

Expected ROUGE scores after training on SAMSum:

Metric	Score
ROUGE-1	~0.42-0.45
ROUGE-2	~0.18-0.21
ROUGE-L	~0.35-0.38

Development

# Install dev dependencies
pip install -r requirements-dev.txt
# Run linting
flake8 src/
black src/ --check
# Format code
black src/
isort src/

🚀 Deployment

Local Development

# Backendcd Text-Summarizer
pip install -r requirements.txt
python app.py
# API available at http://localhost:8000# Frontend (in another terminal)cd Text-Summarizer/frontend
npm install
npm run dev
# Frontend available at http://localhost:5173

Deploy to Render

This project includes a render.yaml Blueprint for easy deployment:

1. Push to GitHub
2. Create Render Account at render.com
3. New Blueprint → Connect your GitHub repo
4. Deploy → Render will auto-detect services

Manual Deployment

Backend (Web Service):

• Build Command: pip install -r requirements.txt
• Start Command: uvicorn app:app --host 0.0.0.0 --port $PORT
• Health Check: /health

Frontend (Static Site):

• Build Command: cd frontend && npm install && npm run build
• Publish Directory: frontend/dist
• Environment: VITE_API_URL=https://your-backend.onrender.com

Environment Variables

Variable	Service	Description
PORT	Backend	Server port (auto-set by Render)
CORS_ORIGINS	Backend	Allowed origins (default: *)
VITE_API_URL	Frontend	Backend API URL

🛠️ Tech Stack

Backend

• Python 3.11 - Core language
• FastAPI - REST API framework
• PyTorch - Deep learning framework
• Transformers - HuggingFace model library
• FLAN-T5 - Pre-trained summarization model

Frontend

• React 19 - UI framework
• Vite - Build tool
• Tailwind CSS - Styling
• Framer Motion - Animations
• GSAP - Advanced animations
• Lenis - Smooth scrolling

📊 Performance

Metric	Score
ROUGE-1	43.53
ROUGE-2	20.01
ROUGE-L	34.78
Inference Time	~3s
Frontend Bundle	453KB (149KB gzipped)

License

MIT License - see LICENSE for details.

👨‍💻 Author

Parshv Patel

• 🎓 UC Berkeley • B.A. Data Science
• 🧠 AI/ML Focus • 4.0 GPA
• 🔗 LinkedIn
• 💻 GitHub
• 🌐 Portfolio