<p>Accurately mining consumer preferences is crucial for enhancing product quality and is an effective strategy for boosting consumer satisfaction. This study proposes a new nonlinear mapping method between sentiment preferences and Kano quality categories of product attributes to guide product improvement. Specifically, with online reviews as data sources, the dictionary-based text sentiment analysis is first used to calculate consumer sentiment polarity. Secondly, a set of functions is designed to illustrate the mapping relationship between sentiment preferences and Kano quality categories. It is designed separately for different Kano quality categories to facilitate the extraction of consumer commonality preferences. Combining attribute importance and preference coefficient, the mining of consumer personality preferences is realized. The Consumer Satisfaction Calculation Model (CSCM) is put forward. Then, to resolve the contradiction between individualization and cost-effectiveness in terms of improvement expenses, a nine-scenario product improvement model (NSPIM) is developed to establish the priority order for enhancing attributes. Finally, the actual data extracted from the Yelp website is utilized to perform simulation experiments in MATLAB. The experimental results help enrich the relevant emotion mining methods and provide scientific guidance for enterprises and product designers to improve products. On the one hand, it helps enterprises to understand consumer preferences and achieve more accurate marketing. On the other hand, it provides the direction for developers to develop effective product improvement programs to improve the core competitiveness of products.</p>