Extracting Consumer Opinion on Indonesian E-Commerce: A Rating Evaluation and Lexicon-Based Sentiment Analysis

Abstract

E-commerce as a business platform offers abundant advantages in modern life all over the world. Sellers and buyers at online marketplaces may get benefits and advantages from e-commerce. One of the advantages is that e-commerce can be accessed anywhere and anytime. Despite providing advantages, e-commerce also has disadvantages including product quality fraud and data theft. Online marketplaces provide facilities for consumer evaluation, through star rating and consumer reviews. In this paper, we focus on the Business-to-Consumer (B2C) e-commerce type and extract consumer opinion data from a leading online marketplace in Indonesia and use text mining approaches to compare the rating evaluation and sentiment analysis on consumer reviews. With 2,937 records, we investigate the relationship between star rating and lexicon-based sentiment analysis. From the results, we found that most consumers do not hesitantly provide a good evaluation indicated by a 5-star rating and positive sentiment of reviews. A quite polarized rating distribution is found and indicates a straightforward consumer opinion. However, a further examination of the relation between rating and review, we discover inconsistencies in consumer opinion where the good rating may also contain negative reviews. Our result findings provide an insight to build a more integrated consumer opinion indicator in e-commerce and that online marketplace sellers need to look deeper at the detailed reviews rating.