Landmark Explanation: An Explainer for Entity Matching Models

Andrea Baraldi, Francesco Del Buono, Matteo Paganelli, Francesco Guerra

Abstract: State-of-the-art approaches model Entity Matching (EM) as a binary classification problem, where Machine (ML) or Deep Learning (DL) based techniques are applied to evaluate if descriptions of pairs of entities refer to the same real-world instance. Despite these approaches have experimentally demonstrated to achieve high effectiveness, their adoption in real scenarios is limited by the lack of interpretability of their behavior. This paper showcases Landmark Explanation1, a tool that makes generic post-hoc (model-agnostic) perturbation-based explanation systems able to explain the behavior of EM models. In particular, Landmark Explanation computes local interpretations, i.e., given a description of a pair of entities and an EM model, it computes the contribution of each term in generating the prediction. The demonstration shows that the explanations generated by Landmark Explanation are effective even for non-matching pairs of entities, a challenge for explanation systems.