Uncertain tourist expressions and data-driven perceived destination image construction: a probabilistic linguistic and multi-criteria decision framework

Understanding how tourists mentally construct destination images from online information has become increasingly important, yet the textual expressions produced by tourists in online reviews often contain vague, uncertain, and multi-interpretable meanings. This study develops an integrated analytical framework to capture and utilize such ambiguity for destination image construction and recommendation. First, probabilistic linguistic term sets (PLTSs) are employed to formally model the uncertain and fuzzy components embedded in tourists' online expressions, enabling a more fine-grained representation of perceived image attributes. Then, latent Dirichlet allocation (LDA) and sentiment analysis are applied to large-scale online reviews to identify thematic structures and emotional orientations underlying tourist discourse. By synthesizing probabilistic linguistic evaluations with extracted topics and sentiments, a multidimensional destination image is constructed that reflects both the semantic diversity and evaluative uncertainty of user-generated content. To transform these insights into decision support, the study further integrates PL-entropy weighting with the PL-MACBETH multi-criteria decision-making approach to generate rational and explainable destination recommendations. A real-world case study verifies the model's effectiveness, demonstrating that the PLTS-based representation significantly improves the interpretation of perceived relevance and importance, and the proposed recommendation mechanism yields consistent and robust decision outcomes. The findings contribute to both tourism research and practice by offering a methodological pathway to better leverage ambiguous online expressions in destination evaluation and recommendation.