All work

Finding discriminatory language in 50,000+ scanned land titles

Problem

Historical land title documents existed only as scanned images, and reviewing them for discriminatory language was a fully manual job.

What I built

An end-to-end pipeline. OpenCV cleans each scanned image, Tesseract OCR extracts the text, and spaCy with scikit-learn classifies passages for review.

Result

Replaced the manual review across 50,000+ documents and saved 400+ hours of work.

Stack

• Python
• OpenCV
• Tesseract OCR
• spaCy
• scikit-learn
• Pandas