Department of Computer, Control and Management Engineering

Topic Modeling is a well-known text-mining strategy that detects potential underlying topics for documents. It plays a pivotal role in recommender systems for processing proliferated user-generated content (UGC) for personalized recommendations. Its application presents unique challenges in tourism sector due to the diversity, dynamicity, and multimodality of tourism data. This study presents a comprehensive analysis of selected promising topic models specifically in context of tourism recommender systems. The study conducts experimental evaluation of models’ performance on five datasets, and highlights their advantages and unique characteristics based on multiple evaluation parameters. Results reveal no best approach in general, rather optimality of models depend on data characteristics, as thoroughly discussed in this paper. It further discusses open issues for the tourism context-related application of topic models, and future research directions.