A Novel Data-Driven Weighted Sentiment Analysis with an Application for Online Medical Review.

Online reviews provide a lot of information for analyzing consumers' satisfaction with products. However, traditional methods analyze overall online reviews, which not only wastes human and material resources, but also produces data analysis deviation. Meanwhile, traditional methods cannot accurately mine the novel features of products after software update. Therefore, a new data-driven method is proposed to overcome the shortcomings of the traditional method. We screen helpful reviews through information entropy to get the product features that customers really care about. We also utilize the uncertainty of information entropy to find the product features that customers follow with interest. Then we obtain the ranking of customer satisfaction with products by weighted sentiment analysis of product features. A case of medical APP is used to verify the availability and effectiveness of the proposed method. The results show that using 56.72% of the original data, 92% of the consistent results can be achieved, and 8% of the novel features can be discovered. Our research method can also be applied to environmental science and other fields. Finally, some interesting conclusions and future research directions are given. [ABSTRACT FROM AUTHOR]