Entity Matching with Quality and Error Guarantees

Abstract

Given two sets of entities X and Y , entity matching aims to decide whether x and y represent the same entity for each pair (x, y) ∈ X × Y . In many scenarios, the only way to ensure perfect accuracy is to launch a costly inspection procedure on every (x, y), whereas performing the procedure |X| · |Y | times is prohibitively expensive. It is, therefore, important to design an algorithm that carries out the procedure on only some pairs, and renders the verdicts on the other pairs automatically with as few mistakes as possible. This article describes an algorithm that achieves the purpose using the methodology of active monotone classification. The algorithm ensures an asymptotically optimal tradeoff between the number of pairs inspected and the number of mistakes made.