Harshness-aware sentiment mining framework for product review. (January 2022)

• Record Type: Journal Article
• Authors: Wang, Xun; Zhou, Ting; Wang, Xiaoyang; Fang, Yili
• Is Part Of: Expert systems with applications. Volume 187(2022)
• Publication Date: 2022-01
• Subjects: Sentiment mining -- Bayesian inference -- Probabilistic graphical model; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2021.115887
• Languages: English
• ISSNs: 0957-4174

Abstract

Abstract: Sentiment mining has been a helpful mechanism that targets to understand the market feedback on certain commodities by utilizing user comments. In general, the process of yielding each comment is essentially associated with his/her criteria for rating (i.e., the degree of harshness ), which makes users provide biased comments. For instance, for a tolerant user, although the user is extremely dissatisfied with the product, harshness still makes her yield a neutral comment which cannot indicate the product quality. Existing work straightforwardly removes the comments of harsh users and those of tolerant ones, which is not the best strategy. To this end, we propose a harshness-aware sentiment analysis framework for product review. First, we depict the process of providing comments from users as a probabilistic graphical model in which the harshness is incorporated. Second, we employ a Bayesian-based inference for sentiment mining. Extensive experimental evaluations have shown that the results of the proposed method are more consistent with the expert evaluations than those of the state-of-the-art methods, and even outperform the method which infers the final evaluations with the ground truth of comments without considering users' harshness. Highlights: We propose a framework of sentiment analysis that well portrays users' harshness. Considering users' harshness can improve the quality of sentiment mining. Experiments show that our method is superior to the state-of-the-art methods. We clarify the reason why the sentiment mining result is far from expert comment.