Projects

Maktaba

Full-text search over 40,000 scanned Arabic manuscripts, with OCR correction built in.

2025 Python, Rust Open source
Visit project Source code
A screenshot of the project goes here.

The problem

Thousands of historical Arabic manuscripts have been scanned, but scans are just pictures. Without searchable text, finding a single passage means reading page after page by eye.

What it does

  • Reads scanned pages and turns them into text.
  • Corrects common recognition mistakes using context.
  • Searches the whole collection in under a second.

How it was built

The recognition pipeline is written in Python. The search index is written in Rust for speed, and the whole system runs on a single small server.