We report on our continuing work on pedigree-based record linkage. In particular, we show how a structured neural network can be designed to learn weights across pieces of information and how the inherent skewness of the data can be reduced by filtering, or blocking, through a series of these networks. The results, both quantitative and qualitative, are… (More)
This paper presents a first attempt at using pedigree-based data to improve record linkage. It describes a composite metric for similarity and a mechanism to extract relevant generational features. Results on a large data set demonstrate promise.
— Genealogical Record Linkage (GRL) is the process of determining whether two pedigrees refer to the same base individual. Unlike other record linkage problems, GRL datasets are extremely sparse and have several multi-valued attributes. In this paper, we describe a metric-based, data mining approach to GRL, and report on its successful application to a… (More)