A Method of Factor Extraction of Sizzle Words Based on Language Resourses

Abstract

“Sizzle” is a major factor in contemporary marketing because it evokes a feeling that indicates customer buying intention and appetite. In recent years, research on the expression of sizzle words has been actively conducted. In this study, we propose a support tool for understanding sizzle words through knowledge extraction by natural language processing. In onomatopoeia research that forms part of the sizzle word, factor analysis is generally used in subject experiments and analysis of questionnaire data. However, generally used factor analysis cannot be applied to language data because of the structure of the frequency matrix. In this study, we apply nonnegative matrix factorization to extract knowledge about sizzle words in review data from a recipe site. The quality of the learning results was improved by weighting the frequency matrix by BM 25. Furthermore, we visualized acquired knowledge of sizzle words using a factor map and word cloud. Experimental results confirmed that keywords of factors influencing the sense of sizzle can be visually grasped.