Exploring Digital Tourism Through Topic Models: A Review and Experimental Study

The surge in the volume and complexity of user-generated content (UGC) and data on digital tourism platforms has raise both opportunities and challenges for its automated analysis. Advanced topic modeling techniques are now necessitated to cater the variety, dynamism, and multifaceted nature of this data, yet their application in digital tourism encounters unique challenges. This study comprehensively reviews prominent and emerging topic models from the categories of probabilistic models, matrix factorization-based models, and neural embedding-based models, describing their systemic architectures and operational mechanisms. In the application context of digital tourism, the study follows an experimental evaluation of the models’ performance on five datasets, across multiple coherence and diversity parameters. Results do not reveal optimality of a single model universally; rather, a model’s effectiveness depends on size and structure characteristics of the data as extensively analyzed in this article. Additionally, the study presents quantitative and qualitative findings, implicit shortcomings along with conclusive deductions, digital tourism application related open issues of topic models, followed by future directions of research.