Towards an Open Source Toolkit for Building Record Linkage Workflows

Record linkage has been subject of research for several decades, and a huge number of record linkage solutions have been proposed, based on probabilistic and empirical paradigms. However, record linkage is a complex process, for the execution of which one single technique is often not enough; it can be seen as composed by distinct phases, each requiring a specific technique and depending on given application and data requirements. Due to such complexity and application dependency, in this paper we propose a toolkit for record linkage, called RELAIS. The toolkit is based on the idea of choosing the most appropriate technique for each phase, and of combining such techniques in a dynamically built record linkage workflow. A real case study validates the RELAIS idea and provides a methodological pattern for driving the design of a record linkage workflow on the basis of the requirements of a real application.