A Product Redesign Approach Based on Negative Text Mining and Kansei Engineering

Abstract

As smart products continue to proliferate, the abundance of negative reviews on online platforms has become a critical source for uncovering product experience deficiencies. However, negative texts often contain implicit emotions, intertwined semantics, and mixed attribute expressions, making traditional analytical methods insufficient for transforming them into design-ready structured knowledge.This study proposes a product redesign framework that integrates negative text mining with Kansei engineering, establishing a complete process from negative affect extraction to visual product optimization. First, a semantic association graph is constructed from review corpora, and a Graph Convolutional Network is employed to identify key negative factors influencing user experience. These negative factors are then mapped to Kansei dimensions through Partial Least Squares Regression, further linked to adjustable design elements to form a quantitative translation pathway from user dissatisfaction to design language.Based on the derived target Kansei directions, a generative design module is introduced to perform guided visual refinement. The generation process is driven by the target affective features and explores multiple feasible visual improvement schemes. The resulting images provide intuitive references for designers, enabling negative user experiences to be directly translated into design decisions.Experimental results demonstrate that the proposed framework effectively identifies deep-seated negative experiences, provides interpretable connections between textual emotions and design characteristics, and supports product form optimization through generative artificial intelligence. This method highlights the value of negative reviews in experience-driven design and offers a systematic approach for emotion-informed improvement of smart products.

Keywords: Negative Text Mining, Kansei Engineering, Graph Convolutional Network, Product Redesign, Generative Design

DOI: 10.54941/ahfe1007430