Published
Proceedings of The 27th World Multi-Conference on Systemics, Cybernetics and Informatics (WMSCI 2023)
Combining Characteristic Information and Part-of-Speech Information in the Classification of a Product Review’s Helpfulness
Abstract
Online shopping has become popular in recent years. In order for shoppers to decide on a product, one often goes for product reviews, written by other customers. Product reviews by customers are considered by some as more genuine and reliable. However, some products have received a large number of reviews, while only a few are usually adequate for a purchase decision. In order to assist potential buyers in selecting only helpful reviews, our study aims to classify a product review as being helpful or unhelpful. It utilizes a review’s part-of-speech (POS) information, as well as characteristic information. While the former can be obtained by analyzing the review’s content, the latter is mostly comprehensible by a quick glance at the review, such as star rating, reviewer’s name and profile, length of the review’s text. The study combines attributes supplied by both types of information and uses them as input to six different classification algorithms. The results show that the combined attribute set improves the classification performance. The improvement agrees for all the six algorithms. The study also compares algorithms by some performance numbers and reports some of the top performers. Moreover, it looks inside some of the models in order to identify some of the most influential attributes.
Combining Characteristic Information and Part-of-Speech Information in the Classification of a Product Review’s Helpfulness. Orlando, Florida, The 27th World Multi-Conference on Systemics, Cybernetics and Informatics (WMSCI 2023) (460-464).