Pictorial Map Generation based on Color Extraction and Sentiment Analysis using SNS Photos

Abstract

In recent years, many SNS photo-posting sites have become popular, and many users share photos of various spots. On the other hand, the top ranking of the purpose of overseas travel as a form of entertainment is related to sentiment such as relaxation and stress reduction. Previous sentiment-based tourist spot recommendations relied heavily on textual data for sentiment analysis. However, they do not utilize the characteristics of photos showing tourist spots. In particular, color information has its specific image and psychological effects, and using this information in recommending tourist information can affect the senses and sentiments of people. For example, the green provides psychological effects, such as relaxation and healing. Therefore, in this work, we focus on the psychological and sentimental effects on colors and propose a method for generating a pictorial map by extracting color information from SNS photos, analyzing the sentiments of colors, and then combining them with the metadata of the SNS photos. It allows users to intuitively search for the desired spots when traveling for purposes related to their sentiments, such as relaxation or stress reduction, by mapping the photos of tourist spots that match their destinations on the map. Furthermore, it is expected to improve sensory behavior, such as choosing colors for a trip. Finally, we developed a visualization interface for the pictorial map using spot information and tag clouds, and we validated the usefulness of the pictorial map visualization by users through a questionnaire survey.