Online reviews have become the modern-day referral, which shapes consumers’ perceptions of products and thus influences product sales performance in the digital economy (Blanco, Sarasa, & Sanclemente, 2010). Prior literature suggests that online consumers' textual information significantly affects product performance and has important strategic value for organizations (Zhou et al., 2018). Sentiment analysis is used to identify the positive and negative tone of textual information (Hu, Bose, Koh, & Liu, 2012) and has become a primary application of analytics when researchers investigate how user-generated information influences product performance. However, most existing online review studies conduct sentiment analysis at the review level, which focuses on identifying the valence of an individual message or review (e.g., Hu et al., 2012; Wu, Huang, & Zhao, 2019), rather than the feature-based, which aims to reveal prior customers’ evaluation of product features in reviews (e.g., Wang, Lu, & Tan, 2018). Since consumers’ fundamental purpose in reading textual reviews is to obtain details about the product attributes’ pros and cons (Xu, 2019), conducting sentiment analysis at review-level fails to measure customer satisfaction concerning each attribute of products or services and does not match the mechanism of how online text reviews are consumed. Therefore, feature-based review-level sentiment analysis better reflects the actual value of textual information in the digital economy.

Abstract Only