Has anyone tried to link and dedupe the various datasets using a probabilistic linkage tool like Splink?

https://moj-analytical-services.github.io/splink/

(Disclaimer: I am the lead author, but the tool is FOSS)