Auto Title – Intelligent Title Generation from Text

---

Overview

Auto Title is a professional, production-ready NLP system that automatically generates concise, accurate, and human-like titles from long-form text such as articles, research papers, blog posts, or reports.

It uses a hybrid AI pipeline combining:

• Abstractive generation (BART + T5)
• Extractive keyphrase fallback (YAKE + TextRank)
• Semantic scoring (Sentence-Transformers)
• Smart post-processing (length, capitalization, clickbait filtering)

This project is a complete, advanced, and professional Python implementation of AutoTitle – Intelligent Title Generation from Text, ready to run as a standalone script or be integrated into a larger project.

---

Features

Feature	Description
Hybrid Generation	Abstractive + Extractive fallback
Semantic Ranking	Uses cosine similarity with lead sentences
Confidence Scoring	Quantifies title quality
Post-Processing	Enforces 5–12 words, title case, removes quotes
Clickbait Filter	Blocks sensationalist phrases
Batch Processing	Generate titles for multiple texts
Lazy Model Loading	Efficient memory usage
CLI & API Ready	Easily integrable into apps

---

Installation

# Clone the repository
git clone https://github.com/aligh993/Auto-Title.git
cd Auto-Title

# Create virtual environment (recommended)
python -m venv venv
source venv/bin/activate  # Linux/Mac
# or
venv\Scripts\activate     # Windows

# Install dependencies
pip install -r requirements.txt

# Download spaCy model
python -m spacy download en_core_web_sm

Requirements

torch>=2.1.0
transformers>=4.40.0
sentence-transformers>=2.5.0
spacy>=3.7.0
yake>=0.4.8
numpy>=1.24.0

Quick Start with Output

python autotitle_advanced.py

Output:

Loading models... (this may take a moment)
AutoTitle ready!

============================================================
 AUTO TITLE GENERATION RESULT 
============================================================
Title:       AI Predicts Alzheimer's 6 Years Early
Method:      Abstractive
Confidence:  0.892
Words:       6
============================================================

Batch Example with Output

texts = [
    "Global temperatures reached record highs in 2024, with July being the hottest month ever recorded.",
    "A new species of deep-sea fish was discovered near the Mariana Trench.",
    "Tesla announced a breakthrough in solid-state battery technology."
]

for i, text in enumerate(texts, 1):
    res = generator.generate(text, max_words=8)
    print(f"{i}. {res['title']} ({res['method']})")

Output:

Batch Processing Example:
1. Global Temperatures Hit Record Highs (extractive)
2. New Deep-Sea Fish Species Discovered (abstractive)
3. Tesla Unveils Solid-State Battery Breakthrough (abstractive)

How It Works

Project Structure

Auto-Title/
├── autotitle_advanced.py    # Main intelligent title generator
├── requirements.txt         # Dependencies
├── README.md                # This file
├── diagram.svg              # Model Diagram
├── data/
│   └── sample_articles.txt  # Test data
└── models/                  # (Optional) Fine-tuned models

Use Cases

• News Aggregators – Auto-title incoming articles
• SEO Optimization – Generate click-worthy blog titles
• Research Papers – Convert abstracts to concise titles
• Email Automation – Smart subject line generation
• Content Curation – Summarize reports and documents
• Document Labeling – Auto-tag PDFs and reports

Contributing

1. Fork the repo
2. Create a branch:

   git checkout -b feature/new-model

3. Commit changes:

   git commit -m "Add T5 refinement layer"

4. Push and open a Pull Request
5. Include tests and update docs

License

This project is licensed under the AGPL-3.0 License. See the LICENSE file for details.

Author

• Ali Ghanbari

• GitHub: @aligh993

• LinkedIn: linkedin.com/in/AliGhanbari

• Email: alighanbari446@gmail.com

One line to capture a thousand words.