Abstract
This software paper introduces the RecordLinkage R package, which implements probabilistic and deterministic methods for linking and deduplicating records across data sources that lack a common, reliable identifier. It provides the full linkage pipeline: generating and comparing candidate record pairs, classifying them as matches or non-matches using weighting and machine-learning methods, and evaluating the results. By packaging these steps as open, reproducible R code, it made principled record linkage accessible to a broad community. Published in The R Journal, RecordLinkage remains a widely used reference implementation, including for privacy-preserving and registry linkage.
record linkage deduplication RecordLinkage R