Date of Award

1-11-2021

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Information Science

First Advisor

John Talburt

Abstract

Since its birth in 1979, e-commerce has grown to become a generation-defining phenomenon. There are millions of platforms from which online transactions can be made, and billions of products available for purchase. As these numbers continue to grow, there is an increased need to exchange, process, and store data about these products. There are also services such as review aggregators and price comparison sites that rely on collecting and integrating data from all over the web. When combining data from multiple sources, it is important to be able to identify which references refer to the same real-life product. Product matching, a variant of Entity Resolution, is the process that aids in solving that challenge. Product matching depends in part on determining how similar a given pair of references are based on their attributes. Similarity measures are utilized to calculate scores indicating the degree of similarity between two attributes. These measures vary in their ability to work with product listings, mainly due to the inconsistent order and redundancy associated with product references. This study seeks to examine the effectiveness of a wide selection of similarity measures for product data, with the goal of determining a “sweet spot” in terms of accuracy and efficiency. The study also proposes the use of string transformations to improve the performance of these similarity measures. The results of this work have the potential to drastically improve the product matching workflow for both practical and research purposes.

Download

Included in

E-Commerce Commons

Share

COinS