Contextual Entity Resolution Approach for Genealogical Data

Due to huge amount of inaccurate information and different types of ambiguity in the available digitized genealogical data, applying Entity Resolution techniques for determining the records referring to the same entity should be considered as the first and still very important step in analysis of this type of data. Traditional methods, use a standard string similarity measure to calculate the similarity among references, neglecting the contextual information available for each reference, and then introduce the most similar pairs as matches. In this paper, first, we introduce a novel blocking strategy to reduce the number of potential candidate pairs. Second, we propose a contextual similarity measure which not only considers the string similarity among references but also contextual information available for them. Third, we evaluate our proposed method extensively from different perspectives and among many discussed patterns, the “early child death” pattern discovered to be prominent.