Topics and Sentiments in Online Place Reviews, an Innovative Way of Understanding the Perception of a City without Asking

Abstract

User-generated content provides rich and easily accessible data for tourism destination managers, especially when combined with a sentiment analysis to uncover perceptions and attitudes. These reviews are often primarily useful in a business/attraction-context and scaling up their relevance for destination management is problematic. Furthermore, the reliability of such online sources can be questioned, thereby impeding its application for research and practice. By combining data of a traditional in-situ survey in five main cultural heritage attraction in Antwerp (Belgium) with scraped data of these same attractions from the TripAdvisor website, this paper attempts to shed a light on the added value and reliability of a big data sentiment analysis. The sentiment analysis combines two lexicons as well as Latent Dirichlet Allocation. The results show promise in that, even though the characteristics between the in-situ sample and the scraped sample are quite different, the sentiments and themes are largely overlapping while the Net Promotor Score as calculated via the TripAdvisor reviews is close to the measured Net Promotor Score through the visitor survey. Still, certain limitations remain within the big data sentiment analysis approach, leading to the conclusion that both methods can be highly compatible in order to efficiently generate deeper, more complete results.