Modern Record Linkage Techniques for a Complex and Ever-changing World

Unleashing the potential of increasingly ubiquitous data to solve grand problems from conservation biology to demography and population estimation often requires linking multiple data sources. Record linkage is the process of resolving duplicates in partially overlapping sets of records from noisy data sources without a unique identifier. The first model-based approach to record linkage was formalized in the seminal 1969 Fellegi and Sunter paper, opening an avenue to automatically link records in a principled fashion. In this proposed session, five approaches to modernize the record linkage model will be presented. Each extends record linkage to novel data contexts. From scaling record linkage to massive scales and efficiently handling ever-changing data, to improving small area estimation and downstream inferential tasks via state-of-the-art approaches to record linkage that preserve differential privacy, each talk will highlight how the established field of record linkage can expand and adapt to the modern world.