Toward human-centered visual design based on sound symbolic information

Novel objects, such as a product with a visual appearance matched with an impression from an object's name, may facilitate the identification of the object and improve user experience. To optimize visual appearance based on an object name, the study proposes a machine learning approach for the automatic prediction of human impressions from word sounds. The proposed method utilized a support vector machine to predict shape labels (i.e., spiky/round/neutral) using acoustic and phonetic features extracted from word sounds. The method demonstrated that acoustic and phonetic features can capture human impressions of shape obtained from word sounds. Moreover, the study conducted a behavioral experiment and demonstrated that shapes matched with impressions from names facilitate recall of the association between names and visual appearance. In addition, the proposed method predicted the degree of facilitation of the memory process. Therefore, the acoustic and phonetic information-based approach exhibited potential for application in the optimization of visual design.