A Latent Dirichlet Model for Unsupervised Entity Resolution

摘要: Entity resolution has received considerable attention in recent years. Given many references to underlying entities, the goal is predict which correspond same entity. We show how extend Latent Dirichlet Allocation model for this task and propose a probabilistic collective entity relational domains where are connected each other. Our approach differs from other recently proposed approaches that it a) generative, b) does not make pair-wise decisions c) captures relations between entities through hidden group variable. novel sampling algorithm unsupervised also takes into account. Additionally, we do assume domain of be known infer number data. demonstrate utility practicality our author two real-world bibliographic datasets. In addition, present preliminary results on characterizing conditions under information useful.