Do travelers' reviews depend on the destination? An analysis in coastal and urban peer‐to‐peer lodgings

Abstract

Our research applies a service, feature-oriented approach to deeply explore the subjective experiences shared publicly by Airbnb guests in their reviews. Our processed data set contains 73,557 reviews of Airbnb stays in coastal and urban destinations between 2017 and 2020. A topic modeling based on the BERTopic approach is applied to detect dense clusters of reviews and identify one highly relevant and interpretable topic per cluster related to core and essential sharing services and surrounding features. Our study, therefore, allows a higher understanding of the relationships between urban versus coastal destinations and guests' preferences. Furthermore, it enables hosts to differentiate the touristic short-rentals lodgings according to customer experiences.

A Topic Modeling Comparison Between LDA, NMF, Top2Vec, and BERTopic to Demystify Twitter Posts

“…First and foremost, compared to other techniques, BERTopic works exceptionally with pretrained embeddings (Sánchez-Franco and Rey-Moreno, 2022) due to a split between clustering the documents and using c-TF-IDF to extract topic representations. Especially owing to the c-TF-IDF procedure (Abuzayed and Al-Khalifa, 2021), BERTopic can support several topic modeling variations, such as guided topic modeling, dynamic topic modeling, or class-based topic modeling.…”

“…Those viewed as the most established, go-to techniques include LDA, latent semantic analysis (LSA), and probabilistic LSA (Albalawi et al, 2020). More recently, however, newly developed algorithms such as NMF, Corex, Top2Vec, and BERTopic have also received, and are continuing to attract, increasing attention from researchers (Obadimu et al, 2019;Sánchez-Franco and Rey-Moreno, 2022). In the social sciences, topic models have formerly been applied to, for example, discover consumers' implicit preferences (Vu et al, 2019;, identify semantic structures on Instagram (Egger and Yu, 2021), and improve recommendation systems (Shafqat and Byun, 2020).…”

“…Regardless of the popularity of LDA within the social science branch, its efficacy in analyzing social media data has been highly criticized (Egger and Yu, 2021;Sánchez-Franco and Rey-Moreno, 2022). In the case of Twitter data, Jaradat and Matskin (2019) argue that, while multiple topics can coexist in a document, LDA tends to neglect co-occurrence relations.…”

“…This research is certainly not without its limitations. While this study responds to a need to utilize Top2Vec and BERTopic for the analysis of short-text data (Egger and Yu, 2021;Sánchez-Franco and Rey-Moreno, 2022), novel language models, such as GPT3 and WuDao 2.0, have continued to emerge as time passes (Nagisetty, 2021), thereby acting as an excellent basis for even more powerful topic modeling approaches. To leverage the use of topic modeling methods, social scientists are encouraged to try and evaluate other newly developed algorithms and to keep their knowledge up to date.…”

“YOU POST, I TRAVEL.” Bloggers' credibility, digital engagement, and travelers' behavioral intention: The mediating role of hedonic and utilitarian motivations

“…Blogs are a type of user‐generated content, which is perceived as more trustworthy than marketing communications (Filieri, 2016), can create expectations about tourism destinations (Narangajavana et al, 2017; Sánchez‐Franco & Rey‐Moreno, 2021), and helps consumers familiarize themselves with the products and services they want to buy (Filieri, 2015). Travel blogs have a strong impact on social media users, who can alter their destination choice if their friends share a negative opinion about it (Lozanov, 2018; Ortaleza & Mangali, 2021).…”