TruthSource

Frontpage
Verification process
Analysis Result 1
Analysis Result 2

Project Story: TruthSource

Inspiration

We created TruthSource after discovering how frequently academic papers misrepresent cited sources. With studies showing up to 21% of citations contain significant errors, these misrepresentations spread through literature like a virus, creating false consensus and misleading researchers who trust citations without verification.

Our Approach

TruthSource combines three powerful technologies:

GROBID: Extracts structured data from academic PDFs
Google's Gemini AI: Provides semantic understanding to verify if citations accurately represent sources
TypeScript: Powers our verification system with confidence scoring

Our modular architecture includes document processing, citation verification, and a flexible database for source papers.

Challenges Overcome

We tackled complex PDF extraction, developed context boundary detection for citations, implemented fuzzy matching for reference variations, and optimized performance for large papers—all while carefully balancing precision and recall in our verification process.

What's Next

We're developing a web interface, academic database integration, a browser extension for real-time verification, and batch processing capabilities for publishers.

TruthSource aims to restore integrity to academic publishing by making citation verification efficient, accessible, and reliable.