Tourism Profiling: A Semi-automatic Classification Model of Points of Interest

Abstract

Reviews are a powerful source of information that helps tourists in their decision-making process. However, using this volume of data to make decisions it is time consuming. For example, the city Foz do Iguaçu, located in Brazil, has more than 44k reviews on TripAdvisor. Based on these opinions, how could a tourist understand if this attraction is good for families, a romantic date, or if it offers a good outdoor experience? Moreover, which other attractions could offer similar experiences? These questions motivated this research, as we try to address the problem of classifying tourism attractions/destinations in profiles. We proposed a hybrid approach, using experts' knowledge and machine-learning with semi-automatic classification models to solve the problem. This paper presents a new approach to classify tourism attractions in profiles using reviews. Our findings show that, the most visited places are not necessarily the most relevant to a specific profile and as such the corresponding group of tourists. Understanding these profiles can aid the discovery or the selection of a travel destination. In addition, it allows governments and the private sector to target tourism marketing actions in the most assertive way.

Keywords

• machine-learning
• reviews
• tourism profiling
• information classification