Which Amazon Reviews Should You Read?
Presentation
Overview
Overview
Description
Customers have to make informed purchasing decisions and online reviews are often a tool customers use to help. However, if there are thousands of reviews, which ones should be read? We present a method of selecting a smaller subset of reviews that is similar to the set of all reviews. This is done through document-term frequency analysis which provides a matrix of term counts for each document (Salton & McGill, 1983) and optimization models including both linear and quadratic models with integer variables. The methods follow closely those found in Kline (1998).
We minimize the distance of the subset from the corpus subject review/word count constraints and also minimize the document/word count subject to distance of the subset from the corpus constraints. The technique is demonstrated on three products with 2000+ reviews from Amazon: a book, an article of clothing, and an electronic device. These products were selected based on each product having a different set (or corpus) of terms used in their reviews. This will test the proposed methods across varying corpora for generalizability.
