Bridging the Trust Gap in the AI Era
Because every claim deserves a source, and every source deserves proper citation.

Why This Matters • Features • Quick Start • Usage • Contributing

We're living in an era where AI can write beautifully, but can't cite properly.

Large Language Models (LLMs) like ChatGPT, Claude, and Gemini are incredible at generating human-like text, but they have a fundamental flaw: they lack reliable citation mechanisms. When an LLM tells you about a scientific study, historical event, or technical concept, you're left wondering:

• 📚 Where did this information come from?
• 🔍 How can I verify these claims?
• 📝 How do I properly cite this in my research?

This creates a trust gap that undermines the reliability of AI-generated content, especially in academic, professional, and research contexts.

Citation Extractor exists to fill this gap.

While LLMs struggle with proper citations, this tool excels at extracting structured, verifiable citation data from any source. It's the missing piece that makes AI-generated content trustworthy and academically sound.

• 📄 PDFs: Academic papers, books, theses, book chapters
• 🌐 Web URLs: Articles, blog posts, online publications
• 🎥 Media Files: Video lectures, podcasts, audio recordings

• Smart Document Classification: Automatically detects if it's a journal article, book, thesis, or book chapter
• Multilingual OCR: Handles English, Chinese (Simplified & Traditional), and more
• Flexible LLM Backend: Works with Ollama (local) or cloud APIs (Gemini, OpenAI)

• CSL-JSON Standard: Compatible with Zotero, Mendeley, EndNote, and all major reference managers
• Multiple Citation Styles: Chicago, APA, MLA, and any CSL style you need
• Structured Metadata: Author, title, publication date, DOI, ISBN, and more

• Smart Page Selection: Processes only the most relevant pages for speed
• Iterative Extraction: Efficiently extracts citation data with early stopping when sufficient information is found
• Offline Processing: Works entirely offline for PDF documents without requiring external API calls
• Batch Processing: Handle multiple documents efficiently

pip install cite-extractor

# Ubuntu/Debian
sudo apt-get install tesseract-ocr mediainfo

# macOS
brew install tesseract mediainfo

# For local LLM support (optional)
# Install Ollama: https://ollama.ai/

# Extract from a PDF
citation "path/to/research-paper.pdf"

# Extract from a URL
citation "https://www.nature.com/articles/s41586-023-06627-7"

# Extract from a video
citation "path/to/conference-talk.mp4"

# Basic usage
citation "document.pdf"

# Specify document type
citation "thesis.pdf" --type thesis

# Use different LLM
citation "paper.pdf" --llm gemini/gemini-1.5-flash

# Custom output directory
citation "book.pdf" --output-dir ./citations

# Specific page range for large documents
citation "book.pdf" --page-range "1-5, -3"

# Different citation style
citation "article.pdf" --citation-style apa

from citation.main import CitationExtractor
from citation.citation_style import format_bibliography

# Initialize with your preferred LLM
extractor = CitationExtractor(llm_model="ollama/qwen3")

# Extract citation data
csl_data = extractor.extract_citation("research-paper.pdf")

if csl_data:
    # Format as bibliography
    bibliography, in_text = format_bibliography([csl_data], "chicago-author-date")
    
    print("📚 Bibliography:")
    print(bibliography)
    
    print("\n📝 In-text citation:")
    print(in_text)

# For non-English documents
citation "chinese-paper.pdf" --lang chi_sim+eng

# Verbose output for debugging
citation "document.pdf" --verbose

# Custom citation style (place .csl file in citation/styles/)
citation "paper.pdf" --citation-style nature

• Automatically cite papers you're reading
• Build bibliographies from PDF collections
• Ensure proper attribution in literature reviews

• Generate citations for thesis references
• Create bibliographies for term papers
• Verify and format existing citations

• Add credible sources to blog posts
• Cite academic backing for claims
• Build trust with properly attributed content

• Add citation capabilities to AI applications
• Verify sources for AI-generated content
• Build trustworthy AI systems

citation "https://arxiv.org/pdf/2301.07041.pdf"

citation "https://www.bbc.com/news/science-environment-64234567"

citation "MIT_6.034_Lecture_1.mp4"

We're thrilled to have you join this mission! 🎉

This project addresses a fundamental need in our AI-driven world, and we believe it can make a real difference in how we handle information credibility. Whether you're a developer, researcher, or just someone who cares about proper attribution, there's a place for you here.

1. 🐛 Report Issues: Found a bug or have a feature request?
2. 💡 Suggest Improvements: Ideas for better citation extraction?
3. 🔧 Submit Code: Bug fixes, new features, or optimizations
4. 📚 Improve Documentation: Help others understand and use the tool
5. 🌍 Add Language Support: Extend OCR and extraction to new languages
6. 🎨 Citation Styles: Add support for more academic citation styles

git clone https://github.com/your-username/citation-extractor.git
cd citation-extractor

# Install development dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Format code
black .

• 🔍 Enhanced Source Detection: Better recognition of document types
• 🌐 Web Scraping: Improved extraction from various websites
• 🎥 Media Support: Better metadata extraction from videos/audio
• 📊 Batch Processing: GUI for handling multiple documents
• 🔗 Integration: Plugins for popular reference managers

This project stands on the shoulders of giants:

• DSPy: For flexible LLM integration
• Tesseract: For OCR capabilities
• citeproc-py: For citation formatting
• The Open Source Community: For making tools like this possible

MIT License - feel free to use this in your projects, commercial or otherwise.

• 📦 PyPI: https://pypi.org/project/cite-extractor/
• 🐛 Issues: Report bugs or request features
• 💬 Discussions: Join the conversation

Made with ❤️ for the research community
Because every claim deserves a source, and every source deserves respect.

⭐ Star this repo if you find it useful! ⭐