Persons Linking in Baptism Records

This paper describes models that can be automatically created from genealogical records (baptisms, marriages and burials). Those
models represent various relationships between people mentioned in the
records. The most important relationship is child - father or mother,
but there can be found others - grandparents, godparents, best men,
midwifes or priests. Usual goal in genealogy is to create models that
are based only on child - parents relationships. Such models are called
family trees. In this paper we describe whole procedure neccessary for
creating such models. We are using rewritten baptisms records of small
village that covers year range between cca 1607 to 1899. Those records
are loaded from simple database and they are transformed to the structure containing all neccessary information about one person. From this
first structure we create another one, that contains all persons mentioned
in the record and which is suitable for comparison. After comparison the
probability that both persons in two records are the same is computed.
If the probability is smaller than threshold, the record is added to the
output database, if it is bigger, it is merged with the second record. Because the rewritten records were hand-connected to the family tree in
genealogical SW and all persons got its own ID, we are then able to find
succes rate of our approach.