Way to go: A graph mining based approach to record linkage for sparse data

Record linkage is an important field for Digital History and Historical Demography, among others. To link data from different recordsets existing algorithms almost always take at least two discriminating data into account, such as a name and date of birth, to link records of an individual. If there is only sparse data available for a person, i.e. only a name, it is usually dismissed as impossible to link by the record linkage literature. This paper introduces an algorithm to link sparse data sets on individuals when, by and large, only nominal data is available.