Examining transaction-specific satisfaction and trust in Airbnb and hotels. An application of BERTopic and Zero-shot text classification

Abstract

With a methodological approach, this article explores the application of data mining to the user-generated content of tourist accommodation on infomediation platforms and social networks. Its objective is to present an algorithm that allows the identification of service characteristics relevant to guest satisfaction and trust. Our study processes unstructured, natural-language data about Airbnb and hotel stays (the final dataset was 12,236 Airbnb sentences and 12,200 hotel sentences from 2018 until September 25 2021). Among the results is a computational algorithm that uses BERTopic to identify latent themes (or topics) in the narratives. Secondly, our analysis applies a Zero-shot classification approach for classifying guest reviews into labels related to guests' satisfaction and trust. Thirdly, we execute a Principal Component Analysis to investigate the sufficiency relationships between extracted topics, customer satisfaction, and trust-based labels. To sum up, and as practical implications, our study adds to the knowledge about the sharing economy by providing insights for developing marketing policies and a better understanding of hospitality services.