Predicting the appreciation of multimodal advertisements

Creativity is an essential factor in successful advertising where catchy and memorable media is produced to persuade the audience. The creative elements in the visual design and in the slogan of an advertisement elevate the overall appeal providing a perceptually grounded attractive message. In this study, we propose the exploitation of creativity cues in textual and visual information for the appreciation prediction of multimodal advertising prints. Moreover, as a novel dimension space of multimodality, we propose using the human sense (i.e., sight, hearing, taste, and smell) information embedded in the language. Our findings show that sensorial information is an invaluable indication of whether the advertisement is appreciated or not. Furthermore, combining linguistic and visual models significantly improves the unimodal appreciation detection performances.