The impact of COVID-19 on the Brazilian food service industry: topic modelling of online reviews

Resumo

The purpose of this paper is twofold. First, it is to discover the underling topics that characterize the customer relationship with restaurants in Brazil. Second, it is to explore the changes in this relationship caused by the necessary actions to mitigate the COVID-19 pandemic. To reach this purpose, 57,381 online reviews of 1,066 restaurants in four Brazilian cities were collected from TripAdvisor. To analyze this data, this research employed text mining and natural language processing techniques, such as topic modelling with Latent Dirichlet Allocation (LDA) and sentiment analysis. Results present a list of the 15 most relevant topics that characterize the customer relationship with restaurants in Brazil. These topics were than divided into five dimensions: Food type, Service, Food quality, Cost benefit, and Place. Due to the COVID-19 pandemic, the number of reviews reduced by 93.27%, comparing with the same period of 2019. In the pandemic period, the four most relevant topics are ‘delivery’, ‘employees’, ‘experience’, and ‘waiter service’. This research is among the first to make use of text mining and topic modelling altogether with sentiment analysis to examine consumers perceptions of Brazilian food service industry, especially to understand the impacts of the COVID-19 pandemic on this sector. The results clearly present the most relevant areas that consumers express interest in restaurants, showing paths for managers facilitate and improve their restaurants operations.