Design and Application of Mapping Model for Font Recommendation System Based on Contents Emotion Analysis

Abstract

Fonts are an important tool that can compensate for the absence of nonverbal and paralinguistic means that are reflected in real-world situations. However, selecting an appropriate font is a process that heavily relies on aesthetic sense and experiential judgment, making it difficult for the general public who are not proficient in using fonts. Therefore, in this study, we intend to implement a service that automatically recommends fonts that match the message when content such as facial expressions and sentences are entered. To this end, we designed an experiment to interpret the emotions associated with different fonts and a model to map the actual content and fonts. In the process of identifying the emotion of the font, We selected emotion keywords to verify the relevance of fonts and quantified their emotional impressions. Since the emotional criteria for content extracted using a deep learning emotional analysis model differed from those for fonts, we devised a new mapping method. We created a mapping model that calculates the correlation between each emotional criterion and determines similarity. We applied this model to confirm the relationship between the emotions of the content and the fonts and developed a system that recommends fonts. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.