Multidimensional Emotional Structures and Intention-Expression Signals in Online Tourism Reviews

초록

Tourism analytics has traditionally relied on sentiment polarity to explain behavioral outcomes, yet emotional expression in online reviews is inherently multidimensional. This study examines how four theoretically grounded emotional dimensions-valence, intensity, authenticity, and affective complexity-are associated with the explicit articulation of revisit intention in review text. Drawing on 7,500 English-language Trip.com reviews (2020-2024), emotional features were extracted using a hybrid NLP pipeline combining fine-tuned BERT classification with lexicon-based arousal scoring and linguistic indices. Revisit intention was operationalized as an explicit intention-expression signal within textual narratives, rather than as a survey-derived attitudinal construct. Logistic regression tested dimensional associations, while K-means clustering identified latent emotional profile types across the corpus. Results indicate that positive valence, greater emotional intensity, and higher affective complexity are each significantly associated with revisit intention expression, whereas emotional authenticity shows no significant direct association. Cluster analysis further reveals distinct emotional configurations that differ systematically in intention-expression prevalence. By reconceptualizing revisit intention as a communicative signal and modeling emotion as a multidimensional structure, this study advances emotion-aware computational analysis in tourism research. The integration of supervised and unsupervised analytics offers a scalable framework for examining how emotional structure shapes behavioral signaling in digital review environments.

키워드