|
Abstract:
|
Since the first edition of Alex Haley's book, Roots, there has been a fascination with tracing family origins. A Google search for "genealogy" results in millions of hits. Some web sites offer users online searches of census index records, which are electronic databases containing summary information from census heads of households. When searching these records, results can be frustrating unless a user enters the exact name, spelling, and age recorded in the census index. This paper describes the implementation of probabilistic record linkage to improve the search. By sampling the 1910 and 1920 census index records, record linkage weights were developed that allow intelligent searches recognizing reasonable variations in search criteria. Using these weights, the chance of missing a record was reduced from more than 50% to less than 2%.
|
