Bayesian Graphical Entity Resolution using Exchangeable Random Partition Priors

Entity resolution (record linkage or deduplication) is the process of identifying and linking duplicate records in databases. In this paper, we propose a Bayesian graphical approach for entity that links to latent entities, where prior representation on structure exchangeable. First, adopt flexible tractable set priors structure, which corresponds special class random partition models. Second, more realistic distortion model categorical/discrete record attributes, corrects logical inconsistency with standard hit-miss model. Third, incorporate hyperpriors improve flexibility. Fourth, employ partially collapsed Gibbs sampler inferential speedups. Using selection private nonprivate data sets, investigate impact our modeling contributions compare two alternative addition, conduct simulation study household survey data, vary distortion, duplication rates size. We find performs consistently than alternatives across variety scenarios typically achieves highest accuracy (F1 score). Open source software available proposed methodology, provide discussion regarding work future directions.