SOARS Text Analysis of Ebay Reviews


Greg Ellison, Department of Statistics, Colorado State University

Friday October 31st, 2014

 

Abstract

Ebay.com is an online auction website that is a rich source of data and complicated user behavior. As well as the ability to rate a transaction as a positive or negative experience, users (both buyers and sellers) have the option of leaving a review describing their transaction. These reviews, while limited to 80 characters are interesting in that a user has the ability to write whatever they want regarding a transaction, and we wish to find relationships between the reviews and selling price, transaction satisfaction, etc.

There are many techniques, both supervised and unsupervised that we wish to explore in analyzing these eBay user reviews. We include common exploratory methods such as wordclouds to display frequency of terms. We use other supervised and unsupervised techniques to answer questions regarding user comments and behavior, and discover relationships between reviews and both transaction ratings and winning bid price.