Exploring product entity matching in a multi-domain landscape - a community detection approach

Abstract

This paper explores entity matching and its vital role in e-commerce to track products across different domains. Focusing on five diverse approaches, we evaluate their performance based on the precision to recall trade-off. Each model is examined theoretically, emphasizing its advantages and limitations. We also discuss potential improvements to enhance their applicability in this field. Our research aims to provide deeper insights into the effectiveness of various entity matching strategies and their performance under distinct priorities, while considering model architecture and room for improvement.