Explainable Error Detection Method for Structured Data using HoloDetect framework

In the past few years, researchers have made significant progress in achieving precise outcomes using error detection tools based on deep learning models. However, many of these models are considered black boxes as they offer limited insights into the reasons and context behind data errors. If we can present the results of these models in a more understandable manner, professionals could utilize them to verify and adjust the model, comprehend the causes of errors, and rectify them accordingly. In this research, we propose a method to generate explanations for the results of HoloDetect framework, which is a few-shot learning framework designed for error detection. We subsequently evaluate the effectiveness of our proposed method using computer-centeric and human-centeric methods.