The Deed Machine

Mapping Prejudice logo

The Deed Machine is a multi-language set of tools that use OCR and crowdsourced transcription to identify racially restrictive covenant language, then map the results.

Racial covenants are clauses that were inserted into property deeds to prevent people who are not white from buying or occupying land. As of June 2024, Mapping Prejudice volunteers have helped to map more than 52,000 covenanted properties across 3 states, with more on the way soon.

The Deed Machine was created at Mapping Prejudice at the University of Minnesota Libraries. Current collaborators include Michael Corey, David Naughton, Tyler Fisher, Katlyn Alapati, Mike Tigas, and the University of Minnesota Libraries IT staff.

A diagram showing the five basic steps of the Deed Machine process: Digitized Docs, Optical Character Recognition, Text, Zooniverse, and Digitized map.

Deed Machine full workflow

A diagram of the components of the Deed Machine. To the left, an initial processing stage using AWS Step Functions is used to output a series of S3 files, which are ingested into a Django project in the center of the diagram.

Indices and tables