A Data Type-Driven Property Alignment Framework for Product Duplicate Detection on the Web

During the last decade daily life has morphed into a world of broadband ubiquity, where devices facilitate constant engagement. As a consequence of this, the area of e-commerce has seen an immense growth. Despite the market opportunities for retailers and the ease for customers to acquire products through webshops, the shift to digital retail has its drawbacks. For example, it leads to cluttered and incomparable information among different webshops, which calls for an automated method to regain homogeneity in product representations. This paper presents a product duplicate detection solution, which exploits a data type-driven property alignment framework. Based on the performed experiment, we show a statistically significant improvement of the F\(_1\)-score from 47.91 % to 78.13 % compared to an existing state-of-the-art approach.