Abstract
In recent years, it has become necessary to establish methods for quantitatively evaluating human Kansei and impressions to meet the needs of consumers. In this study, a method for quantitatively evaluating impressions of apparel products using online reviews is proposed. Four impression scales (“Dark-Bright,” “Tight-Loose,” “Formal-Casual,” and “Simple-Fancy”) were defined, and an impression dictionary was constructed that includes impression words containing degree adverbs, extreme impression words, and negation words. Validation using large-scale data confirmed that negation words accurately captured impression polarity, while degree adverbs and extreme impression words reflected impression intensity. Statistical analysis showed that scales related to physical characteristics (such as color and fit) were sensitive to modifiers, whereas the Simple-Fancy scale remained stable. In the future, model accuracy will be further validated and the applicability of the method to other fields will be explored.

This publication has 15 references indexed in Scilit: