How are Tourists Different? - Reading Geo-tagged Photos through a Deep Learning Model

Abstract

With the popularity of social media, it is possible to use UGC (user-generated content) for tourist behavior study. However, current UGC-based tourism research is still text-centric; the visual content of user-generated photos is rarely used for tourism destination due to a technology restriction. Aiming to mitigate this deficiency and built a cornerstone for a diversified utilization of UGC, this study innovatively explored the visual content of geo-attached photos with the help of a deep learning model. Setting “How are tourists different?” as the topic, it analyzed 29081 photos shared by tourists from Europe, North America, and Asia in Hong Kong. The chi-square test and frequency analysis were used to compare the perceived differences among the three continents. The density analysis tool of ArcGIS was used to compare the behavior differences from the spatial dimension. Theoretically, this study tested the feasibility of interdisciplinary technology in analyzing the visual content of tourists’ photos for tourism research. Practically, this research provides pictorial pieces of evidence for the market promotion of tourism destination image.