Identifying sentiment-dependent product features from online reviews

Abstract

This paper presents a method to correlate relevant product features to the sales rank data. Instead of going through the labor-intensive surveys, online product reviews have become an efficient source to gather consumer preferences. The contribution of the paper is to relate the content of reviews to a product’s sales rank that implicitly reflects the motivation behind what drives customers to purchase the product. After using part-of-speech tagging to extract the relevant feature and opinion pairs from the reviews, the extracted data along with the review ratings and price become the variables to explain the sales rank. An experiment is run for wearable technology products to illustrate the methodology and interpret the result. The result indicates that the positive opinion for battery and negative opinion for sleep tracker are significant towards sales rank, while price is not.